Nina Odoux

NOTEBOOK 2

Este cuaderno se enfocara en data splitting y EDA analisis del data set

1. VERIFICAR BALANCE
2. SPLIT TRAIN/TEST
3. CATEGORICAL AND CONTINUOUS
4. UNIVARIABLE ANALISIS
5. CONCLUSION UNIVARIABLE

---

• For each type CATEGORICAL AND CONTINUOUS analysis of:

6. OUTLIERS
7. MISSING VALUES
8. CORRELATION

#importacion de librerias :
import os
import pandas as pd 
import numpy as np   
import matplotlib.pyplot as plt 
import seaborn as sns      
from sklearn.model_selection import train_test_split
import time
import gc
import psutil
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import StandardScaler
from sklearn.impute import KNNImputer


#importar funciones de mi archivo ipynb:
import sys
sys.path.append("../src")  
import eda_utils as eda

seed = 123
pd.set_option('display.max_columns', 123)
pd.set_option('display.max_rows', 5000)

Lectura de datos del preprocesado inicial

data = pd.read_csv("../data/preprocessed_data/01_preprocessed_data.csv")
data

	SK_ID_CURR	TARGET	NAME_CONTRACT_TYPE	CODE_GENDER	FLAG_OWN_CAR	FLAG_OWN_REALTY	CNT_CHILDREN	AMT_INCOME_TOTAL	AMT_CREDIT	AMT_ANNUITY	AMT_GOODS_PRICE	NAME_TYPE_SUITE	NAME_INCOME_TYPE	NAME_EDUCATION_TYPE	NAME_FAMILY_STATUS	NAME_HOUSING_TYPE	REGION_POPULATION_RELATIVE	DAYS_BIRTH	DAYS_EMPLOYED	DAYS_REGISTRATION	DAYS_ID_PUBLISH	OWN_CAR_AGE	FLAG_MOBIL	FLAG_EMP_PHONE	FLAG_WORK_PHONE	FLAG_CONT_MOBILE	FLAG_PHONE	FLAG_EMAIL	OCCUPATION_TYPE	CNT_FAM_MEMBERS	REGION_RATING_CLIENT	REGION_RATING_CLIENT_W_CITY	WEEKDAY_APPR_PROCESS_START	HOUR_APPR_PROCESS_START	REG_REGION_NOT_LIVE_REGION	REG_REGION_NOT_WORK_REGION	LIVE_REGION_NOT_WORK_REGION	REG_CITY_NOT_LIVE_CITY	REG_CITY_NOT_WORK_CITY	LIVE_CITY_NOT_WORK_CITY	ORGANIZATION_TYPE	EXT_SOURCE_1	EXT_SOURCE_2	EXT_SOURCE_3	APARTMENTS_AVG	BASEMENTAREA_AVG	YEARS_BEGINEXPLUATATION_AVG	YEARS_BUILD_AVG	COMMONAREA_AVG	ELEVATORS_AVG	ENTRANCES_AVG	FLOORSMAX_AVG	FLOORSMIN_AVG	LANDAREA_AVG	LIVINGAPARTMENTS_AVG	LIVINGAREA_AVG	NONLIVINGAPARTMENTS_AVG	NONLIVINGAREA_AVG	APARTMENTS_MODE	BASEMENTAREA_MODE	YEARS_BEGINEXPLUATATION_MODE	YEARS_BUILD_MODE	COMMONAREA_MODE	ELEVATORS_MODE	ENTRANCES_MODE	FLOORSMAX_MODE	FLOORSMIN_MODE	LANDAREA_MODE	LIVINGAPARTMENTS_MODE	LIVINGAREA_MODE	NONLIVINGAPARTMENTS_MODE	NONLIVINGAREA_MODE	APARTMENTS_MEDI	BASEMENTAREA_MEDI	YEARS_BEGINEXPLUATATION_MEDI	YEARS_BUILD_MEDI	COMMONAREA_MEDI	ELEVATORS_MEDI	ENTRANCES_MEDI	FLOORSMAX_MEDI	FLOORSMIN_MEDI	LANDAREA_MEDI	LIVINGAPARTMENTS_MEDI	LIVINGAREA_MEDI	NONLIVINGAPARTMENTS_MEDI	NONLIVINGAREA_MEDI	FONDKAPREMONT_MODE	HOUSETYPE_MODE	TOTALAREA_MODE	WALLSMATERIAL_MODE	EMERGENCYSTATE_MODE	DAYS_LAST_PHONE_CHANGE	FLAG_DOCUMENT_2	FLAG_DOCUMENT_3	FLAG_DOCUMENT_4	FLAG_DOCUMENT_5	FLAG_DOCUMENT_6	FLAG_DOCUMENT_7	FLAG_DOCUMENT_8	FLAG_DOCUMENT_9	FLAG_DOCUMENT_10	FLAG_DOCUMENT_11	FLAG_DOCUMENT_12	FLAG_DOCUMENT_13	FLAG_DOCUMENT_14	FLAG_DOCUMENT_15	FLAG_DOCUMENT_16	FLAG_DOCUMENT_17	FLAG_DOCUMENT_18	FLAG_DOCUMENT_19	FLAG_DOCUMENT_20	FLAG_DOCUMENT_21	AMT_REQ_CREDIT_BUREAU_HOUR	AMT_REQ_CREDIT_BUREAU_DAY	AMT_REQ_CREDIT_BUREAU_WEEK	AMT_REQ_CREDIT_BUREAU_MON	AMT_REQ_CREDIT_BUREAU_QRT	AMT_REQ_CREDIT_BUREAU_YEAR
0	100002	1	CASH LOANS	MALE	NO	YES	0	202500.0	406597.5	24700.5	351000.0	UNACCOMPANIED	WORKING	SECONDARY / SECONDARY SPECIAL	SINGLE / NOT MARRIED	HOUSE / APARTMENT	0.018801	-9461	-637	-3648.0	-2120	NaN	1	1	0	1	1	0	LABORERS	1.0	2	2	WEDNESDAY	10	0	0	0	0	0	0	BUSINESS ENTITY TYPE 3	0.083037	0.262949	0.139376	0.0247	0.0369	0.9722	0.6192	0.0143	0.00	0.0690	0.0833	0.1250	0.0369	0.0202	0.0190	0.0000	0.0000	0.0252	0.0383	0.9722	0.6341	0.0144	0.0000	0.0690	0.0833	0.1250	0.0377	0.0220	0.0198	0.0	0.0000	0.0250	0.0369	0.9722	0.6243	0.0144	0.00	0.0690	0.0833	0.1250	0.0375	0.0205	0.0193	0.0000	0.0000	REG OPER ACCOUNT	BLOCK OF FLATS	0.0149	STONE, BRICK	NO	-1134.0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.0	0.0	0.0	0.0	0.0	1.0
1	100003	0	CASH LOANS	FEMALE	NO	NO	0	270000.0	1293502.5	35698.5	1129500.0	FAMILY	STATE SERVANT	HIGHER EDUCATION	MARRIED	HOUSE / APARTMENT	0.003541	-16765	-1188	-1186.0	-291	NaN	1	1	0	1	1	0	CORE STAFF	2.0	1	1	MONDAY	11	0	0	0	0	0	0	SCHOOL	0.311267	0.622246	NaN	0.0959	0.0529	0.9851	0.7960	0.0605	0.08	0.0345	0.2917	0.3333	0.0130	0.0773	0.0549	0.0039	0.0098	0.0924	0.0538	0.9851	0.8040	0.0497	0.0806	0.0345	0.2917	0.3333	0.0128	0.0790	0.0554	0.0	0.0000	0.0968	0.0529	0.9851	0.7987	0.0608	0.08	0.0345	0.2917	0.3333	0.0132	0.0787	0.0558	0.0039	0.0100	REG OPER ACCOUNT	BLOCK OF FLATS	0.0714	BLOCK	NO	-828.0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.0	0.0	0.0	0.0	0.0	0.0
2	100004	0	REVOLVING LOANS	MALE	YES	YES	0	67500.0	135000.0	6750.0	135000.0	UNACCOMPANIED	WORKING	SECONDARY / SECONDARY SPECIAL	SINGLE / NOT MARRIED	HOUSE / APARTMENT	0.010032	-19046	-225	-4260.0	-2531	26.0	1	1	1	1	1	0	LABORERS	1.0	2	2	MONDAY	9	0	0	0	0	0	0	GOVERNMENT	NaN	0.555912	0.729567	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	-815.0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.0	0.0	0.0	0.0	0.0	0.0
3	100006	0	CASH LOANS	FEMALE	NO	YES	0	135000.0	312682.5	29686.5	297000.0	UNACCOMPANIED	WORKING	SECONDARY / SECONDARY SPECIAL	CIVIL MARRIAGE	HOUSE / APARTMENT	0.008019	-19005	-3039	-9833.0	-2437	NaN	1	1	0	1	0	0	LABORERS	2.0	2	2	WEDNESDAY	17	0	0	0	0	0	0	BUSINESS ENTITY TYPE 3	NaN	0.650442	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	-617.0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	NaN	NaN	NaN	NaN	NaN	NaN
4	100007	0	CASH LOANS	MALE	NO	YES	0	121500.0	513000.0	21865.5	513000.0	UNACCOMPANIED	WORKING	SECONDARY / SECONDARY SPECIAL	SINGLE / NOT MARRIED	HOUSE / APARTMENT	0.028663	-19932	-3038	-4311.0	-3458	NaN	1	1	0	1	0	0	CORE STAFF	1.0	2	2	THURSDAY	11	0	0	0	0	1	1	RELIGION	NaN	0.322738	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	-1106.0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.0	0.0	0.0	0.0	0.0	0.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
307506	456251	0	CASH LOANS	MALE	NO	NO	0	157500.0	254700.0	27558.0	225000.0	UNACCOMPANIED	WORKING	SECONDARY / SECONDARY SPECIAL	SEPARATED	WITH PARENTS	0.032561	-9327	-236	-8456.0	-1982	NaN	1	1	0	1	0	0	SALES STAFF	1.0	1	1	THURSDAY	15	0	0	0	0	0	0	SERVICES	0.145570	0.681632	NaN	0.2021	0.0887	0.9876	0.8300	0.0202	0.22	0.1034	0.6042	0.2708	0.0594	0.1484	0.1965	0.0753	0.1095	0.1008	0.0172	0.9782	0.7125	0.0172	0.0806	0.0345	0.4583	0.0417	0.0094	0.0882	0.0853	0.0	0.0125	0.2040	0.0887	0.9876	0.8323	0.0203	0.22	0.1034	0.6042	0.2708	0.0605	0.1509	0.2001	0.0757	0.1118	REG OPER ACCOUNT	BLOCK OF FLATS	0.2898	STONE, BRICK	NO	-273.0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	NaN	NaN	NaN	NaN	NaN	NaN
307507	456252	0	CASH LOANS	FEMALE	NO	YES	0	72000.0	269550.0	12001.5	225000.0	UNACCOMPANIED	PENSIONER	SECONDARY / SECONDARY SPECIAL	WIDOW	HOUSE / APARTMENT	0.025164	-20775	365243	-4388.0	-4090	NaN	1	0	0	1	1	0	NaN	1.0	2	2	MONDAY	8	0	0	0	0	0	0	XNA	NaN	0.115992	NaN	0.0247	0.0435	0.9727	0.6260	0.0022	0.00	0.1034	0.0833	0.1250	0.0579	0.0202	0.0257	0.0000	0.0000	0.0252	0.0451	0.9727	0.6406	0.0022	0.0000	0.1034	0.0833	0.1250	0.0592	0.0220	0.0267	0.0	0.0000	0.0250	0.0435	0.9727	0.6310	0.0022	0.00	0.1034	0.0833	0.1250	0.0589	0.0205	0.0261	0.0000	0.0000	REG OPER ACCOUNT	BLOCK OF FLATS	0.0214	STONE, BRICK	NO	0.0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	NaN	NaN	NaN	NaN	NaN	NaN
307508	456253	0	CASH LOANS	FEMALE	NO	YES	0	153000.0	677664.0	29979.0	585000.0	UNACCOMPANIED	WORKING	HIGHER EDUCATION	SEPARATED	HOUSE / APARTMENT	0.005002	-14966	-7921	-6737.0	-5150	NaN	1	1	0	1	0	1	MANAGERS	1.0	3	3	THURSDAY	9	0	0	0	0	1	1	SCHOOL	0.744026	0.535722	0.218859	0.1031	0.0862	0.9816	0.7484	0.0123	0.00	0.2069	0.1667	0.2083	NaN	0.0841	0.9279	0.0000	0.0000	0.1050	0.0894	0.9816	0.7583	0.0124	0.0000	0.2069	0.1667	0.2083	NaN	0.0918	0.9667	0.0	0.0000	0.1041	0.0862	0.9816	0.7518	0.0124	0.00	0.2069	0.1667	0.2083	NaN	0.0855	0.9445	0.0000	0.0000	REG OPER ACCOUNT	BLOCK OF FLATS	0.7970	PANEL	NO	-1909.0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1.0	0.0	0.0	1.0	0.0	1.0
307509	456254	1	CASH LOANS	FEMALE	NO	YES	0	171000.0	370107.0	20205.0	319500.0	UNACCOMPANIED	COMMERCIAL ASSOCIATE	SECONDARY / SECONDARY SPECIAL	MARRIED	HOUSE / APARTMENT	0.005313	-11961	-4786	-2562.0	-931	NaN	1	1	0	1	0	0	LABORERS	2.0	2	2	WEDNESDAY	9	0	0	0	1	1	0	BUSINESS ENTITY TYPE 1	NaN	0.514163	0.661024	0.0124	NaN	0.9771	NaN	NaN	NaN	0.0690	0.0417	NaN	NaN	NaN	0.0061	NaN	NaN	0.0126	NaN	0.9772	NaN	NaN	NaN	0.0690	0.0417	NaN	NaN	NaN	0.0063	NaN	NaN	0.0125	NaN	0.9771	NaN	NaN	NaN	0.0690	0.0417	NaN	NaN	NaN	0.0062	NaN	NaN	NaN	BLOCK OF FLATS	0.0086	STONE, BRICK	NO	-322.0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.0	0.0	0.0	0.0	0.0	0.0
307510	456255	0	CASH LOANS	FEMALE	NO	NO	0	157500.0	675000.0	49117.5	675000.0	UNACCOMPANIED	COMMERCIAL ASSOCIATE	HIGHER EDUCATION	MARRIED	HOUSE / APARTMENT	0.046220	-16856	-1262	-5128.0	-410	NaN	1	1	1	1	1	0	LABORERS	2.0	1	1	THURSDAY	20	0	0	0	0	1	1	BUSINESS ENTITY TYPE 3	0.734460	0.708569	0.113922	0.0742	0.0526	0.9881	NaN	0.0176	0.08	0.0690	0.3750	NaN	NaN	NaN	0.0791	NaN	0.0000	0.0756	0.0546	0.9881	NaN	0.0178	0.0806	0.0690	0.3750	NaN	NaN	NaN	0.0824	NaN	0.0000	0.0749	0.0526	0.9881	NaN	0.0177	0.08	0.0690	0.3750	NaN	NaN	NaN	0.0805	NaN	0.0000	NaN	BLOCK OF FLATS	0.0718	PANEL	NO	-787.0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.0	0.0	0.0	2.0	0.0	1.0

307511 rows × 118 columns

VERIFICAR BALANCE

Los datos no parecen seguir un orden específico que impacte la repartición durante el proceso de división. Sin embargo, se observa un desequilibrio en las proporciones de las clases de TARGET que debe ser tratado. Para abordar el desequilibrio de clases, podríamos considerar métodos como el re-sampling, el uso de pesos de clase o realizar un muestreo estratificado para asegurar una representación equilibrada de cada clase durante el entrenamiento del modelo.

STRATIFICATION

Asegura que la distribución de clases de la variable objetivo se mantenga tanto en el TRAIN como en el TEST. Esto es importante cuando se trabajan con clases desequilibradas, ya que previene que el conjunto de TEST tenga una distribución desigual de las clases en comparación con el conjunto de datos original.

SEPARACION TRAIN Y TEST

porque permite evaluar cómo de bien el modelo generalizara a datos no vistos previamente y se adapta a nuevos datos desconocidos

#features seran (X), que son independientes y target variable sera (y): dependiente en teoria de las variables de X :

#removar nuestra variable objetivo del conjunto X porque es la que es dependiente:
X = data.drop('TARGET', axis=1)  

#asignar y como variable objetivo TARGET:
y = data['TARGET']

#stratified train-test split (80% train, 20% test), enfocado en una stratificacion de la variable TARGET:
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, stratify=y, random_state=seed
)


data_test = pd.concat([X_test, y_test],axis=1)
data_train = pd.concat([X_train, y_train],axis=1)

# verificacion de la distribucion de las clases en los diferentes conjuntos:
print("Original class distribution:")
print(y.value_counts(normalize=True))  # Proportion in original data
print("\nTrain set class distribution:")
print(y_train.value_counts(normalize=True))  # Proportion in train set
print("\nTest set class distribution:")
print(y_test.value_counts(normalize=True))  # Proportion in test set

Original class distribution:
0    0.919271
1    0.080729
Name: TARGET, dtype: float64

Train set class distribution:
0    0.919271
1    0.080729
Name: TARGET, dtype: float64

Test set class distribution:
0    0.919272
1    0.080728
Name: TARGET, dtype: float64

Stratification permitio que las distribuciones de las clases conserven el mismo patron como lo del dataset de origen

• Eran alrededor del 92% para class 0 y un 8% para class 1

Visualización descriptiva de los datos

• valores nulos por filas
• por columnas

Por columnas (variable)

#cqlcular el % de missing por columna
missing_col = eda.check_missing_values(data_train)

# crear un DataFrame con las columnas ordenadas por % de missing en orden DESC:
missing_col_sorted = missing_col.sort_values(ascending=False)

#reset el index
missing_col_sorted = missing_col_sorted.reset_index()
print(missing_col_sorted)

% of NAN per column:
 CODE_GENDER                      0.001219
AMT_ANNUITY                      0.004471
AMT_GOODS_PRICE                  0.085770
NAME_TYPE_SUITE                  0.407710
OWN_CAR_AGE                     65.985253
OCCUPATION_TYPE                 31.276219
CNT_FAM_MEMBERS                  0.000813
EXT_SOURCE_1                    56.310364
EXT_SOURCE_2                     0.209749
EXT_SOURCE_3                    19.907483
APARTMENTS_AVG                  50.715424
BASEMENTAREA_AVG                58.485903
YEARS_BEGINEXPLUATATION_AVG     48.752480
YEARS_BUILD_AVG                 66.479139
COMMONAREA_AVG                  69.847322
ELEVATORS_AVG                   53.250707
ENTRANCES_AVG                   50.298771
FLOORSMAX_AVG                   49.723993
FLOORSMIN_AVG                   67.831534
LANDAREA_AVG                    59.359452
LIVINGAPARTMENTS_AVG            68.351842
LIVINGAREA_AVG                  50.143085
NONLIVINGAPARTMENTS_AVG         69.427823
NONLIVINGAREA_AVG               55.127069
APARTMENTS_MODE                 50.715424
BASEMENTAREA_MODE               58.485903
YEARS_BEGINEXPLUATATION_MODE    48.752480
YEARS_BUILD_MODE                66.479139
COMMONAREA_MODE                 69.847322
ELEVATORS_MODE                  53.250707
ENTRANCES_MODE                  50.298771
FLOORSMAX_MODE                  49.723993
FLOORSMIN_MODE                  67.831534
LANDAREA_MODE                   59.359452
LIVINGAPARTMENTS_MODE           68.351842
LIVINGAREA_MODE                 50.143085
NONLIVINGAPARTMENTS_MODE        69.427823
NONLIVINGAREA_MODE              55.127069
APARTMENTS_MEDI                 50.715424
BASEMENTAREA_MEDI               58.485903
YEARS_BEGINEXPLUATATION_MEDI    48.752480
YEARS_BUILD_MEDI                66.479139
COMMONAREA_MEDI                 69.847322
ELEVATORS_MEDI                  53.250707
ENTRANCES_MEDI                  50.298771
FLOORSMAX_MEDI                  49.723993
FLOORSMIN_MEDI                  67.831534
LANDAREA_MEDI                   59.359452
LIVINGAPARTMENTS_MEDI           68.351842
LIVINGAREA_MEDI                 50.143085
NONLIVINGAPARTMENTS_MEDI        69.427823
NONLIVINGAREA_MEDI              55.127069
FONDKAPREMONT_MODE              68.368102
HOUSETYPE_MODE                  50.147556
TOTALAREA_MODE                  48.227293
WALLSMATERIAL_MODE              50.813388
EMERGENCYSTATE_MODE             47.361468
AMT_REQ_CREDIT_BUREAU_HOUR      13.573136
AMT_REQ_CREDIT_BUREAU_DAY       13.573136
AMT_REQ_CREDIT_BUREAU_WEEK      13.573136
AMT_REQ_CREDIT_BUREAU_MON       13.573136
AMT_REQ_CREDIT_BUREAU_QRT       13.573136
AMT_REQ_CREDIT_BUREAU_YEAR      13.573136
dtype: float64
                           index          0
0                COMMONAREA_MODE  69.847322
1                COMMONAREA_MEDI  69.847322
2                 COMMONAREA_AVG  69.847322
3       NONLIVINGAPARTMENTS_MODE  69.427823
4        NONLIVINGAPARTMENTS_AVG  69.427823
5       NONLIVINGAPARTMENTS_MEDI  69.427823
6             FONDKAPREMONT_MODE  68.368102
7           LIVINGAPARTMENTS_AVG  68.351842
8          LIVINGAPARTMENTS_MODE  68.351842
9          LIVINGAPARTMENTS_MEDI  68.351842
10                FLOORSMIN_MEDI  67.831534
11                FLOORSMIN_MODE  67.831534
12                 FLOORSMIN_AVG  67.831534
13              YEARS_BUILD_MODE  66.479139
14              YEARS_BUILD_MEDI  66.479139
15               YEARS_BUILD_AVG  66.479139
16                   OWN_CAR_AGE  65.985253
17                 LANDAREA_MEDI  59.359452
18                 LANDAREA_MODE  59.359452
19                  LANDAREA_AVG  59.359452
20             BASEMENTAREA_MEDI  58.485903
21              BASEMENTAREA_AVG  58.485903
22             BASEMENTAREA_MODE  58.485903
23                  EXT_SOURCE_1  56.310364
24            NONLIVINGAREA_MODE  55.127069
25             NONLIVINGAREA_AVG  55.127069
26            NONLIVINGAREA_MEDI  55.127069
27                 ELEVATORS_AVG  53.250707
28                ELEVATORS_MEDI  53.250707
29                ELEVATORS_MODE  53.250707
30            WALLSMATERIAL_MODE  50.813388
31                APARTMENTS_AVG  50.715424
32               APARTMENTS_MODE  50.715424
33               APARTMENTS_MEDI  50.715424
34                ENTRANCES_MODE  50.298771
35                ENTRANCES_MEDI  50.298771
36                 ENTRANCES_AVG  50.298771
37                HOUSETYPE_MODE  50.147556
38               LIVINGAREA_MEDI  50.143085
39               LIVINGAREA_MODE  50.143085
40                LIVINGAREA_AVG  50.143085
41                FLOORSMAX_MEDI  49.723993
42                FLOORSMAX_MODE  49.723993
43                 FLOORSMAX_AVG  49.723993
44   YEARS_BEGINEXPLUATATION_AVG  48.752480
45  YEARS_BEGINEXPLUATATION_MEDI  48.752480
46  YEARS_BEGINEXPLUATATION_MODE  48.752480
47                TOTALAREA_MODE  48.227293
48           EMERGENCYSTATE_MODE  47.361468
49               OCCUPATION_TYPE  31.276219
50                  EXT_SOURCE_3  19.907483
51     AMT_REQ_CREDIT_BUREAU_DAY  13.573136
52    AMT_REQ_CREDIT_BUREAU_WEEK  13.573136
53     AMT_REQ_CREDIT_BUREAU_MON  13.573136
54     AMT_REQ_CREDIT_BUREAU_QRT  13.573136
55    AMT_REQ_CREDIT_BUREAU_HOUR  13.573136
56    AMT_REQ_CREDIT_BUREAU_YEAR  13.573136
57               NAME_TYPE_SUITE   0.407710
58                  EXT_SOURCE_2   0.209749
59               AMT_GOODS_PRICE   0.085770
60                   AMT_ANNUITY   0.004471
61                   CODE_GENDER   0.001219
62               CNT_FAM_MEMBERS   0.000813

eda.plot_missing_values(data)

#Valores faltantes por fila y con variable objetivo para cada fila 
missing_row_sorted = eda.check_missing_per_row(data_train)

#Visualizacion del dataframe :
missing_row_sorted

	missing_percentage	TARGET
236260	50.000000	0
249616	50.000000	1
224619	49.152542	0
267335	49.152542	0
26398	49.152542	0
...	...	...
33699	0.000000	0
239409	0.000000	0
221021	0.000000	0
167597	0.000000	0
183048	0.000000	0

246008 rows × 2 columns

VISUALIZATION OF CONTINUOUS AND CATEGORICAL VARIABLES

#llamo a la funcion de mi archivo de funciones: 
#con esta funcion, pusé en una lista las categoricas: categort, object, y numericos con menos de 20 valores unicos 

#y en una otra lista las variables continuas que son tipo float o int, con mas de 20 valores unicos 


categorical_vars, continuous_vars = eda.dame_variables_categoricas(data_train)

print("Variables categóricas:", categorical_vars)
print("----------------------------------------------------------------")
print("Variables continuas:", continuous_vars)

Variables categóricas: ['NAME_CONTRACT_TYPE', 'CODE_GENDER', 'FLAG_OWN_CAR', 'FLAG_OWN_REALTY', 'CNT_CHILDREN', 'NAME_TYPE_SUITE', 'NAME_INCOME_TYPE', 'NAME_EDUCATION_TYPE', 'NAME_FAMILY_STATUS', 'NAME_HOUSING_TYPE', 'FLAG_MOBIL', 'FLAG_EMP_PHONE', 'FLAG_WORK_PHONE', 'FLAG_CONT_MOBILE', 'FLAG_PHONE', 'FLAG_EMAIL', 'OCCUPATION_TYPE', 'CNT_FAM_MEMBERS', 'REGION_RATING_CLIENT', 'REGION_RATING_CLIENT_W_CITY', 'WEEKDAY_APPR_PROCESS_START', 'REG_REGION_NOT_LIVE_REGION', 'REG_REGION_NOT_WORK_REGION', 'LIVE_REGION_NOT_WORK_REGION', 'REG_CITY_NOT_LIVE_CITY', 'REG_CITY_NOT_WORK_CITY', 'LIVE_CITY_NOT_WORK_CITY', 'ORGANIZATION_TYPE', 'FONDKAPREMONT_MODE', 'HOUSETYPE_MODE', 'WALLSMATERIAL_MODE', 'EMERGENCYSTATE_MODE', 'FLAG_DOCUMENT_2', 'FLAG_DOCUMENT_3', 'FLAG_DOCUMENT_4', 'FLAG_DOCUMENT_5', 'FLAG_DOCUMENT_6', 'FLAG_DOCUMENT_7', 'FLAG_DOCUMENT_8', 'FLAG_DOCUMENT_9', 'FLAG_DOCUMENT_10', 'FLAG_DOCUMENT_11', 'FLAG_DOCUMENT_12', 'FLAG_DOCUMENT_13', 'FLAG_DOCUMENT_14', 'FLAG_DOCUMENT_15', 'FLAG_DOCUMENT_16', 'FLAG_DOCUMENT_17', 'FLAG_DOCUMENT_18', 'FLAG_DOCUMENT_19', 'FLAG_DOCUMENT_20', 'FLAG_DOCUMENT_21', 'AMT_REQ_CREDIT_BUREAU_HOUR', 'AMT_REQ_CREDIT_BUREAU_DAY', 'AMT_REQ_CREDIT_BUREAU_WEEK', 'AMT_REQ_CREDIT_BUREAU_QRT', 'TARGET']
----------------------------------------------------------------
Variables continuas: ['SK_ID_CURR', 'AMT_INCOME_TOTAL', 'AMT_CREDIT', 'AMT_ANNUITY', 'AMT_GOODS_PRICE', 'REGION_POPULATION_RELATIVE', 'DAYS_BIRTH', 'DAYS_EMPLOYED', 'DAYS_REGISTRATION', 'DAYS_ID_PUBLISH', 'OWN_CAR_AGE', 'HOUR_APPR_PROCESS_START', 'EXT_SOURCE_1', 'EXT_SOURCE_2', 'EXT_SOURCE_3', 'APARTMENTS_AVG', 'BASEMENTAREA_AVG', 'YEARS_BEGINEXPLUATATION_AVG', 'YEARS_BUILD_AVG', 'COMMONAREA_AVG', 'ELEVATORS_AVG', 'ENTRANCES_AVG', 'FLOORSMAX_AVG', 'FLOORSMIN_AVG', 'LANDAREA_AVG', 'LIVINGAPARTMENTS_AVG', 'LIVINGAREA_AVG', 'NONLIVINGAPARTMENTS_AVG', 'NONLIVINGAREA_AVG', 'APARTMENTS_MODE', 'BASEMENTAREA_MODE', 'YEARS_BEGINEXPLUATATION_MODE', 'YEARS_BUILD_MODE', 'COMMONAREA_MODE', 'ELEVATORS_MODE', 'ENTRANCES_MODE', 'FLOORSMAX_MODE', 'FLOORSMIN_MODE', 'LANDAREA_MODE', 'LIVINGAPARTMENTS_MODE', 'LIVINGAREA_MODE', 'NONLIVINGAPARTMENTS_MODE', 'NONLIVINGAREA_MODE', 'APARTMENTS_MEDI', 'BASEMENTAREA_MEDI', 'YEARS_BEGINEXPLUATATION_MEDI', 'YEARS_BUILD_MEDI', 'COMMONAREA_MEDI', 'ELEVATORS_MEDI', 'ENTRANCES_MEDI', 'FLOORSMAX_MEDI', 'FLOORSMIN_MEDI', 'LANDAREA_MEDI', 'LIVINGAPARTMENTS_MEDI', 'LIVINGAREA_MEDI', 'NONLIVINGAPARTMENTS_MEDI', 'NONLIVINGAREA_MEDI', 'TOTALAREA_MODE', 'DAYS_LAST_PHONE_CHANGE', 'AMT_REQ_CREDIT_BUREAU_MON', 'AMT_REQ_CREDIT_BUREAU_YEAR']

#creo carpeta de imagenes dedicada a la EDA
os.makedirs('../images/02_notebook_images', exist_ok=True)

#llamo mi funcion para visualizar las categoricas, y respeto a la variable TARGET:
eda.plot_all_variables(data_train, categorical_vars, 'TARGET')

Analyzing NAME_CONTRACT_TYPE

Analyzing CODE_GENDER

Analyzing FLAG_OWN_CAR

Analyzing FLAG_OWN_REALTY

Analyzing CNT_CHILDREN

Analyzing NAME_TYPE_SUITE

Analyzing NAME_INCOME_TYPE

Analyzing NAME_EDUCATION_TYPE

Analyzing NAME_FAMILY_STATUS

Analyzing NAME_HOUSING_TYPE

Analyzing FLAG_MOBIL

Analyzing FLAG_EMP_PHONE

Analyzing FLAG_WORK_PHONE

Analyzing FLAG_CONT_MOBILE

Analyzing FLAG_PHONE

Analyzing FLAG_EMAIL

Analyzing OCCUPATION_TYPE

Analyzing CNT_FAM_MEMBERS

Analyzing REGION_RATING_CLIENT

Analyzing REGION_RATING_CLIENT_W_CITY

Analyzing WEEKDAY_APPR_PROCESS_START

Analyzing REG_REGION_NOT_LIVE_REGION

Analyzing REG_REGION_NOT_WORK_REGION

Analyzing LIVE_REGION_NOT_WORK_REGION

Analyzing REG_CITY_NOT_LIVE_CITY

Analyzing REG_CITY_NOT_WORK_CITY

Analyzing LIVE_CITY_NOT_WORK_CITY

Analyzing ORGANIZATION_TYPE

Analyzing FONDKAPREMONT_MODE

Analyzing HOUSETYPE_MODE

Analyzing WALLSMATERIAL_MODE

Analyzing EMERGENCYSTATE_MODE

Analyzing FLAG_DOCUMENT_2

Analyzing FLAG_DOCUMENT_3

Analyzing FLAG_DOCUMENT_4

Analyzing FLAG_DOCUMENT_5

Analyzing FLAG_DOCUMENT_6

Analyzing FLAG_DOCUMENT_7

Analyzing FLAG_DOCUMENT_8

Analyzing FLAG_DOCUMENT_9

Analyzing FLAG_DOCUMENT_10

Analyzing FLAG_DOCUMENT_11

Analyzing FLAG_DOCUMENT_12

Analyzing FLAG_DOCUMENT_13

Analyzing FLAG_DOCUMENT_14

Analyzing FLAG_DOCUMENT_15

Analyzing FLAG_DOCUMENT_16

Analyzing FLAG_DOCUMENT_17

Analyzing FLAG_DOCUMENT_18

Analyzing FLAG_DOCUMENT_19

Analyzing FLAG_DOCUMENT_20

Analyzing FLAG_DOCUMENT_21

Analyzing AMT_REQ_CREDIT_BUREAU_HOUR

Analyzing AMT_REQ_CREDIT_BUREAU_DAY

Analyzing AMT_REQ_CREDIT_BUREAU_WEEK

Analyzing AMT_REQ_CREDIT_BUREAU_QRT

Analyzing TARGET

# llamo la funcion que grafica, guarda las imagenes de mis variables continuas:
eda.plot_all_features(data_train, continuous_vars, target_col='TARGET')

Iniciando análisis de 61 características continuas

Procesando lote 1 de 21

Gráfico de 'SK_ID_CURR' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'AMT_INCOME_TOTAL' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'AMT_CREDIT' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 2 de 21

Gráfico de 'AMT_ANNUITY' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'AMT_GOODS_PRICE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'REGION_POPULATION_RELATIVE' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 3 de 21

Gráfico de 'DAYS_BIRTH' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'DAYS_EMPLOYED' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'DAYS_REGISTRATION' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 4 de 21

Gráfico de 'DAYS_ID_PUBLISH' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'OWN_CAR_AGE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'HOUR_APPR_PROCESS_START' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 5 de 21

Gráfico de 'EXT_SOURCE_1' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'EXT_SOURCE_2' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'EXT_SOURCE_3' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 6 de 21

Gráfico de 'APARTMENTS_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'BASEMENTAREA_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'YEARS_BEGINEXPLUATATION_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 7 de 21

Gráfico de 'YEARS_BUILD_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'COMMONAREA_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'ELEVATORS_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 8 de 21

Gráfico de 'ENTRANCES_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'FLOORSMAX_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'FLOORSMIN_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 9 de 21

Gráfico de 'LANDAREA_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'LIVINGAPARTMENTS_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'LIVINGAREA_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 10 de 21

Gráfico de 'NONLIVINGAPARTMENTS_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'NONLIVINGAREA_AVG' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'APARTMENTS_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 11 de 21

Gráfico de 'BASEMENTAREA_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'YEARS_BEGINEXPLUATATION_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'YEARS_BUILD_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 12 de 21

Gráfico de 'COMMONAREA_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'ELEVATORS_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'ENTRANCES_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 13 de 21

Gráfico de 'FLOORSMAX_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'FLOORSMIN_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'LANDAREA_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 14 de 21

Gráfico de 'LIVINGAPARTMENTS_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'LIVINGAREA_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'NONLIVINGAPARTMENTS_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 15 de 21

Gráfico de 'NONLIVINGAREA_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'APARTMENTS_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'BASEMENTAREA_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 16 de 21

Gráfico de 'YEARS_BEGINEXPLUATATION_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'YEARS_BUILD_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'COMMONAREA_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 17 de 21

Gráfico de 'ELEVATORS_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'ENTRANCES_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'FLOORSMAX_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 18 de 21

Gráfico de 'FLOORSMIN_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'LANDAREA_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'LIVINGAPARTMENTS_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 19 de 21

Gráfico de 'LIVINGAREA_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'NONLIVINGAPARTMENTS_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'NONLIVINGAREA_MEDI' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 20 de 21

Gráfico de 'TOTALAREA_MODE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'DAYS_LAST_PHONE_CHANGE' guardado exitosamente en '..\images\02_notebook_images'.

Gráfico de 'AMT_REQ_CREDIT_BUREAU_MON' guardado exitosamente en '..\images\02_notebook_images'.

Procesando lote 21 de 21

Gráfico de 'AMT_REQ_CREDIT_BUREAU_YEAR' guardado exitosamente en '..\images\02_notebook_images'.

Proceso completado exitosamente.

Conclusión

Para concluir:

Vamos a destacar las tendencias observadas en la distribución de las variables en relación con TARGET para identificar aquellas que, a simple vista, parecen tener mayor influencia en la clasificación.

Factores laborales

• ORGANIZATION_TYPE muestra que ciertos tipos de trabajos, como limpieza, electricidad y cultura, están más asociados con la clase 1, lo que sugiere una relación entre el tipo de ámbito laboral y la propensión a retrasos.
• OCCUPATION_TYPE refuerza esta tendencia, destacando que los trabajos no calificados, como camareros, conductores, cocineros y personal de seguridad, tienen una mayor proporción en la clase 1.

Situación socioeconómica

• En NAME_INCOME_TYPE, los desempleados y quienes están en permiso por maternidad tienen una mayor probabilidad de retraso (clase 1), reflejando factores socioeconómicos que notamos en la vida real.
• Los propietarios de una vivienda y un coche tienden a pertenecer más a la clase 0, aunque esta diferencia no es significativa.

Factores demográficos

• En cuanto al género, los hombres están más representados en la clase 1, mientras que las mujeres predominan en la clase 0.
• La educación parece ser crucial: un mayor nivel educativo está asociado con la clase 0, mientras que niveles bajos se relacionan con la clase 1.
• La edad también influye: los clientes más mayores tienden a estar en la clase 0, mientras que los más jóvenes son más propensos a retrasos (clase 1).

Características del préstamo

• Los préstamos revolving están mayormente asociados con la clase 0, mientras que los préstamos cash predominan en la clase 1. Esto podría deberse a que los préstamos revolving reflejan una mejor gestión financiera.
• Un crédito más alto parece estar vinculado a la clase 0, posiblemente debido a filtros más estrictos para aprobar montos elevados.
• Aunque se esperaría que las annuities más altas se relacionen con la clase 1, los datos muestran que están más presentes en la clase 0, con algunos outliers en la clase 1.

Factores familiares

• El número de ninos impacta significativamente: a medida que aumenta, también lo hace la proporción de clientes en la clase 1. significando una mayor probabilidad de retraso.
• Sin embargo, no se observan patrones claros relacionados con el estado civil segun si el cliente que pidio el prestamo estaba casado o no.

Factores externos

• Las clasificaciones externas (no tenemos insights sobre como se hizo la clasificacion), como el rating de región y rating de cliente, indican que regiones y clientes de tipo 3 tienen mayor probabilidad de pertenecer a la clase 1. Sugeriendo que mas alto el rating, mas riesgos surgen

Conclusion para la variable INCOME

• El ingreso del cliente es una variable que considero imprescindible,

Se observa una tendencia central más alta en la clase 0, que presenta una mediana de ingresos superior. La dispersión es prácticamente igual, dado que el rango intercuartil es bastante similar en ambas clases. Aunque hay una ligera diferencia de valores extremos en la clase 0 (significando que gente con mas ingresos tiene mas probabilidad de ser de clase 0), no es lo suficientemente significativa como para concluir que las personas con retraso tienen una diferencia considerable en los ingresos en comparación con otros casos.

A continuación, se tratan los valores missing, las correlaciones de las variables continuas y los outliers

continuous_vars

['SK_ID_CURR',
 'AMT_INCOME_TOTAL',
 'AMT_CREDIT',
 'AMT_ANNUITY',
 'AMT_GOODS_PRICE',
 'REGION_POPULATION_RELATIVE',
 'DAYS_BIRTH',
 'DAYS_EMPLOYED',
 'DAYS_REGISTRATION',
 'DAYS_ID_PUBLISH',
 'OWN_CAR_AGE',
 'HOUR_APPR_PROCESS_START',
 'EXT_SOURCE_1',
 'EXT_SOURCE_2',
 'EXT_SOURCE_3',
 'APARTMENTS_AVG',
 'BASEMENTAREA_AVG',
 'YEARS_BEGINEXPLUATATION_AVG',
 'YEARS_BUILD_AVG',
 'COMMONAREA_AVG',
 'ELEVATORS_AVG',
 'ENTRANCES_AVG',
 'FLOORSMAX_AVG',
 'FLOORSMIN_AVG',
 'LANDAREA_AVG',
 'LIVINGAPARTMENTS_AVG',
 'LIVINGAREA_AVG',
 'NONLIVINGAPARTMENTS_AVG',
 'NONLIVINGAREA_AVG',
 'APARTMENTS_MODE',
 'BASEMENTAREA_MODE',
 'YEARS_BEGINEXPLUATATION_MODE',
 'YEARS_BUILD_MODE',
 'COMMONAREA_MODE',
 'ELEVATORS_MODE',
 'ENTRANCES_MODE',
 'FLOORSMAX_MODE',
 'FLOORSMIN_MODE',
 'LANDAREA_MODE',
 'LIVINGAPARTMENTS_MODE',
 'LIVINGAREA_MODE',
 'NONLIVINGAPARTMENTS_MODE',
 'NONLIVINGAREA_MODE',
 'APARTMENTS_MEDI',
 'BASEMENTAREA_MEDI',
 'YEARS_BEGINEXPLUATATION_MEDI',
 'YEARS_BUILD_MEDI',
 'COMMONAREA_MEDI',
 'ELEVATORS_MEDI',
 'ENTRANCES_MEDI',
 'FLOORSMAX_MEDI',
 'FLOORSMIN_MEDI',
 'LANDAREA_MEDI',
 'LIVINGAPARTMENTS_MEDI',
 'LIVINGAREA_MEDI',
 'NONLIVINGAPARTMENTS_MEDI',
 'NONLIVINGAREA_MEDI',
 'TOTALAREA_MODE',
 'DAYS_LAST_PHONE_CHANGE',
 'AMT_REQ_CREDIT_BUREAU_MON',
 'AMT_REQ_CREDIT_BUREAU_YEAR']

OUTLIERS

#identificar los valores extremos para variables continuas 
# basándose en un rango definido por un intervalo de confianza multiplicado 
# por una desviación estándar
eda.get_deviation_of_mean_perc(data_train, continuous_vars, target='TARGET', multiplier=3)

	0.0	1.0	variable	sum_outlier_values	porcentaje_sum_null_values
0	0.938095	0.061905	AMT_INCOME_TOTAL	210	0.000854
1	0.958988	0.041012	AMT_CREDIT	2609	0.010605
2	0.960951	0.039049	AMT_ANNUITY	2356	0.009577
3	0.959916	0.040084	AMT_GOODS_PRICE	3318	0.013487
4	0.959733	0.040267	REGION_POPULATION_RELATIVE	6730	0.027357
5	0.958403	0.041597	DAYS_REGISTRATION	601	0.002443
6	0.916479	0.083521	OWN_CAR_AGE	2670	0.010853
7	0.897638	0.102362	HOUR_APPR_PROCESS_START	508	0.002065
8	0.950358	0.049642	APARTMENTS_AVG	2377	0.009662
9	0.949077	0.050923	BASEMENTAREA_AVG	1571	0.006386
10	0.917431	0.082569	YEARS_BEGINEXPLUATATION_AVG	545	0.002215
11	0.921466	0.078534	YEARS_BUILD_AVG	955	0.003882
12	0.944767	0.055233	COMMONAREA_AVG	1376	0.005593
13	0.953775	0.046225	ELEVATORS_AVG	1947	0.007914
14	0.938833	0.061167	ENTRANCES_AVG	1782	0.007244
15	0.957355	0.042645	FLOORSMAX_AVG	2087	0.008483
16	0.970588	0.029412	FLOORSMIN_AVG	476	0.001935
17	0.935252	0.064748	LANDAREA_AVG	1668	0.006780
18	0.948313	0.051687	LIVINGAPARTMENTS_AVG	1393	0.005662
19	0.947904	0.052096	LIVINGAREA_AVG	2553	0.010378
20	0.932743	0.067257	NONLIVINGAPARTMENTS_AVG	565	0.002297
21	0.948454	0.051546	NONLIVINGAREA_AVG	1940	0.007886
22	0.948505	0.051495	APARTMENTS_MODE	2408	0.009788
23	0.944745	0.055255	BASEMENTAREA_MODE	1665	0.006768
24	0.915414	0.084586	YEARS_BEGINEXPLUATATION_MODE	532	0.002163
25	0.922441	0.077559	YEARS_BUILD_MODE	967	0.003931
26	0.939839	0.060161	COMMONAREA_MODE	1363	0.005540
27	0.949329	0.050671	ELEVATORS_MODE	2684	0.010910
28	0.938244	0.061756	ENTRANCES_MODE	1765	0.007175
29	0.958768	0.041232	FLOORSMAX_MODE	2110	0.008577
30	0.971429	0.028571	FLOORSMIN_MODE	385	0.001565
31	0.935972	0.064028	LANDAREA_MODE	1718	0.006984
32	0.944095	0.055905	LIVINGAPARTMENTS_MODE	1431	0.005817
33	0.945673	0.054327	LIVINGAREA_MODE	2669	0.010849
34	0.926554	0.073446	NONLIVINGAPARTMENTS_MODE	531	0.002158
35	0.949058	0.050942	NONLIVINGAREA_MODE	1963	0.007979
36	0.949979	0.050021	APARTMENTS_MEDI	2419	0.009833
37	0.947937	0.052063	BASEMENTAREA_MEDI	1575	0.006402
38	0.913386	0.086614	YEARS_BEGINEXPLUATATION_MEDI	508	0.002065
39	0.922280	0.077720	YEARS_BUILD_MEDI	965	0.003923
40	0.943844	0.056156	COMMONAREA_MEDI	1389	0.005646
41	0.953981	0.046019	ELEVATORS_MEDI	1934	0.007862
42	0.938582	0.061418	ENTRANCES_MEDI	1791	0.007280
43	0.957515	0.042485	FLOORSMAX_MEDI	2189	0.008898
44	0.970852	0.029148	FLOORSMIN_MEDI	446	0.001813
45	0.938596	0.061404	LANDAREA_MEDI	1710	0.006951
46	0.947820	0.052180	LIVINGAPARTMENTS_MEDI	1399	0.005687
47	0.948968	0.051032	LIVINGAREA_MEDI	2567	0.010435
48	0.930728	0.069272	NONLIVINGAPARTMENTS_MEDI	563	0.002289
49	0.948692	0.051308	NONLIVINGAREA_MEDI	1949	0.007923
50	0.956044	0.043956	TOTALAREA_MODE	2639	0.010727
51	0.965974	0.034026	DAYS_LAST_PHONE_CHANGE	529	0.002150
52	0.948478	0.051522	AMT_REQ_CREDIT_BUREAU_MON	2562	0.010414
53	0.911004	0.088996	AMT_REQ_CREDIT_BUREAU_YEAR	2708	0.011008

El método de la desviación estándar es un método clásico que he utilizado y que parece adecuado para la distribución de mis datos. Me parece confiable, no decidi explorar métodos alternativos para obtener una mayor precisión.

Para concluir el analisis de valores extremos continuos:

Una proporción tan baja de valores atípicos globalmente significa que su impacto global en la distribución probablemente sea limitado.
Además, en variables importantes como el ingreso (income), el número de valores atípicos es extremadamente bajo.
No es indispensable tratarlos de momento, pero quedo atenta a la eleccion del modelo para ver si se debe ajustar la precision y si puede tener una importancia adaptarlos si el modelo lo pide.

Se podria utilizar métodos como capping, la imputación con la mediana o la media, las transformaciones matemáticas (logaritmo, raíz cuadrada), o removarlos...

CORRELATION CATEGORICAL AND NUMERICAL SEPARATELY

• MATRIZ Y COEF. DE CORRELACION

eda.get_corr_matrix(dataset = data_train[continuous_vars], 
                metodo='pearson', size_figure=[10,8])

0

corr = data_train[continuous_vars].corr('pearson')
new_corr = corr.abs()
new_corr.loc[:,:] = np.tril(new_corr, k=-1) # below main lower triangle of an array
new_corr = new_corr.stack().to_frame('correlation').reset_index().sort_values(by='correlation', ascending=False)
new_corr[new_corr['correlation']>0.55]

	level_0	level_1	correlation
2824	YEARS_BUILD_MEDI	YEARS_BUILD_AVG	0.998419
3134	FLOORSMIN_MEDI	FLOORSMIN_AVG	0.997352
3072	FLOORSMAX_MEDI	FLOORSMAX_AVG	0.997141
3010	ENTRANCES_MEDI	ENTRANCES_AVG	0.996948
2948	ELEVATORS_MEDI	ELEVATORS_AVG	0.996008
2886	COMMONAREA_MEDI	COMMONAREA_AVG	0.995816
3320	LIVINGAREA_MEDI	LIVINGAREA_AVG	0.995450
2638	APARTMENTS_MEDI	APARTMENTS_AVG	0.995270
3258	LIVINGAPARTMENTS_MEDI	LIVINGAPARTMENTS_AVG	0.994466
2700	BASEMENTAREA_MEDI	BASEMENTAREA_AVG	0.994035
2762	YEARS_BEGINEXPLUATATION_MEDI	YEARS_BEGINEXPLUATATION_AVG	0.993125
3444	NONLIVINGAREA_MEDI	NONLIVINGAREA_AVG	0.991649
3196	LANDAREA_MEDI	LANDAREA_AVG	0.991599
3382	NONLIVINGAPARTMENTS_MEDI	NONLIVINGAPARTMENTS_AVG	0.990498
1970	YEARS_BUILD_MODE	YEARS_BUILD_AVG	0.989127
2838	YEARS_BUILD_MEDI	YEARS_BUILD_MODE	0.989118
3148	FLOORSMIN_MEDI	FLOORSMIN_MODE	0.988433
3086	FLOORSMAX_MEDI	FLOORSMAX_MODE	0.988204
246	AMT_GOODS_PRICE	AMT_CREDIT	0.986997
2280	FLOORSMIN_MODE	FLOORSMIN_AVG	0.986046
2218	FLOORSMAX_MODE	FLOORSMAX_AVG	0.985710
2962	ELEVATORS_MEDI	ELEVATORS_MODE	0.982707
3210	LANDAREA_MEDI	LANDAREA_MODE	0.980788
3024	ENTRANCES_MEDI	ENTRANCES_MODE	0.980273
2900	COMMONAREA_MEDI	COMMONAREA_MODE	0.979032
2094	ELEVATORS_MODE	ELEVATORS_AVG	0.978604
2156	ENTRANCES_MODE	ENTRANCES_AVG	0.977388
2652	APARTMENTS_MEDI	APARTMENTS_MODE	0.977085
3396	NONLIVINGAPARTMENTS_MEDI	NONLIVINGAPARTMENTS_MODE	0.977053
2714	BASEMENTAREA_MEDI	BASEMENTAREA_MODE	0.976871
2032	COMMONAREA_MODE	COMMONAREA_AVG	0.976245
3272	LIVINGAPARTMENTS_MEDI	LIVINGAPARTMENTS_MODE	0.975746
3458	NONLIVINGAREA_MEDI	NONLIVINGAREA_MODE	0.975426
3334	LIVINGAREA_MEDI	LIVINGAREA_MODE	0.974965
1784	APARTMENTS_MODE	APARTMENTS_AVG	0.973294
2342	LANDAREA_MODE	LANDAREA_AVG	0.973156
2466	LIVINGAREA_MODE	LIVINGAREA_AVG	0.972183
1846	BASEMENTAREA_MODE	BASEMENTAREA_AVG	0.972080
2404	LIVINGAPARTMENTS_MODE	LIVINGAPARTMENTS_AVG	0.970693
1908	YEARS_BEGINEXPLUATATION_MODE	YEARS_BEGINEXPLUATATION_AVG	0.970327
2528	NONLIVINGAPARTMENTS_MODE	NONLIVINGAPARTMENTS_AVG	0.967621
2590	NONLIVINGAREA_MODE	NONLIVINGAREA_AVG	0.967063
2776	YEARS_BEGINEXPLUATATION_MEDI	YEARS_BEGINEXPLUATATION_MODE	0.960662
1540	LIVINGAPARTMENTS_AVG	APARTMENTS_AVG	0.944185
3248	LIVINGAPARTMENTS_MEDI	APARTMENTS_AVG	0.942998
3276	LIVINGAPARTMENTS_MEDI	APARTMENTS_MEDI	0.942776
2408	LIVINGAPARTMENTS_MODE	APARTMENTS_MODE	0.939285
2648	APARTMENTS_MEDI	LIVINGAPARTMENTS_AVG	0.936461
2662	APARTMENTS_MEDI	LIVINGAPARTMENTS_MODE	0.933201
2394	LIVINGAPARTMENTS_MODE	APARTMENTS_AVG	0.931936
3503	TOTALAREA_MODE	LIVINGAREA_AVG	0.925681
3531	TOTALAREA_MODE	LIVINGAREA_MEDI	0.920420
3337	LIVINGAREA_MEDI	APARTMENTS_MEDI	0.917647
3262	LIVINGAPARTMENTS_MEDI	APARTMENTS_MODE	0.915566
1601	LIVINGAREA_AVG	APARTMENTS_AVG	0.915396
3309	LIVINGAREA_MEDI	APARTMENTS_AVG	0.914270
2649	APARTMENTS_MEDI	LIVINGAREA_AVG	0.914240
2469	LIVINGAREA_MODE	APARTMENTS_MODE	0.912045
1794	APARTMENTS_MODE	LIVINGAPARTMENTS_AVG	0.909813
3517	TOTALAREA_MODE	LIVINGAREA_MODE	0.900182
2663	APARTMENTS_MEDI	LIVINGAREA_MODE	0.897528
3323	LIVINGAREA_MEDI	APARTMENTS_MODE	0.895802
2455	LIVINGAREA_MODE	APARTMENTS_AVG	0.895123
3492	TOTALAREA_MODE	APARTMENTS_AVG	0.893126
1795	APARTMENTS_MODE	LIVINGAREA_AVG	0.892349
3520	TOTALAREA_MODE	APARTMENTS_MEDI	0.887156
3347	LIVINGAREA_MEDI	LIVINGAPARTMENTS_MEDI	0.884883
3259	LIVINGAPARTMENTS_MEDI	LIVINGAREA_AVG	0.883318
1611	LIVINGAREA_AVG	LIVINGAPARTMENTS_AVG	0.881072
2479	LIVINGAREA_MODE	LIVINGAPARTMENTS_MODE	0.879649
3319	LIVINGAREA_MEDI	LIVINGAPARTMENTS_AVG	0.879001
3333	LIVINGAREA_MEDI	LIVINGAPARTMENTS_MODE	0.874785
2405	LIVINGAPARTMENTS_MODE	LIVINGAREA_AVG	0.873256
3342	LIVINGAREA_MEDI	ELEVATORS_MEDI	0.869387
1606	LIVINGAREA_AVG	ELEVATORS_AVG	0.868331
2954	ELEVATORS_MEDI	LIVINGAREA_AVG	0.866723
3314	LIVINGAREA_MEDI	ELEVATORS_AVG	0.866426
3506	TOTALAREA_MODE	APARTMENTS_MODE	0.864362
3273	LIVINGAPARTMENTS_MEDI	LIVINGAREA_MODE	0.858390
2474	LIVINGAREA_MODE	ELEVATORS_MODE	0.856916
3328	LIVINGAREA_MEDI	ELEVATORS_MODE	0.856738
2100	ELEVATORS_MODE	LIVINGAREA_AVG	0.853372
2465	LIVINGAREA_MODE	LIVINGAPARTMENTS_AVG	0.853118
3502	TOTALAREA_MODE	LIVINGAPARTMENTS_AVG	0.849229
3530	TOTALAREA_MODE	LIVINGAPARTMENTS_MEDI	0.847481
3497	TOTALAREA_MODE	ELEVATORS_AVG	0.843365
2968	ELEVATORS_MEDI	LIVINGAREA_MODE	0.841639
2460	LIVINGAREA_MODE	ELEVATORS_AVG	0.839344
3525	TOTALAREA_MODE	ELEVATORS_MEDI	0.837275
2971	ELEVATORS_MEDI	APARTMENTS_MEDI	0.836541
1235	ELEVATORS_AVG	APARTMENTS_AVG	0.836042
3516	TOTALAREA_MODE	LIVINGAPARTMENTS_MODE	0.835947
2943	ELEVATORS_MEDI	APARTMENTS_AVG	0.834482
2643	APARTMENTS_MEDI	ELEVATORS_AVG	0.833536
2103	ELEVATORS_MODE	APARTMENTS_MODE	0.825541
2657	APARTMENTS_MEDI	ELEVATORS_MODE	0.824718
2089	ELEVATORS_MODE	APARTMENTS_AVG	0.821834
3511	TOTALAREA_MODE	ELEVATORS_MODE	0.820205
3281	LIVINGAPARTMENTS_MEDI	ELEVATORS_MEDI	0.812438
3253	LIVINGAPARTMENTS_MEDI	ELEVATORS_AVG	0.810733
1545	LIVINGAPARTMENTS_AVG	ELEVATORS_AVG	0.810014
2953	ELEVATORS_MEDI	LIVINGAPARTMENTS_AVG	0.807994
2957	ELEVATORS_MEDI	APARTMENTS_MODE	0.807919
2413	LIVINGAPARTMENTS_MODE	ELEVATORS_MODE	0.806637
1789	APARTMENTS_MODE	ELEVATORS_AVG	0.804933
3267	LIVINGAPARTMENTS_MEDI	ELEVATORS_MODE	0.798001
2967	ELEVATORS_MEDI	LIVINGAPARTMENTS_MODE	0.797162
2399	LIVINGAPARTMENTS_MODE	ELEVATORS_AVG	0.795324
2099	ELEVATORS_MODE	LIVINGAPARTMENTS_AVG	0.793141
247	AMT_GOODS_PRICE	AMT_ANNUITY	0.775701
185	AMT_ANNUITY	AMT_CREDIT	0.770691
1425	FLOORSMIN_AVG	FLOORSMAX_AVG	0.742447
3161	FLOORSMIN_MEDI	FLOORSMAX_MEDI	0.740578
3133	FLOORSMIN_MEDI	FLOORSMAX_AVG	0.740053
3073	FLOORSMAX_MEDI	FLOORSMIN_AVG	0.740023
3147	FLOORSMIN_MEDI	FLOORSMAX_MODE	0.730022
2219	FLOORSMAX_MODE	FLOORSMIN_AVG	0.729350
2293	FLOORSMIN_MODE	FLOORSMAX_MODE	0.727067
3087	FLOORSMAX_MEDI	FLOORSMIN_MODE	0.723708
2279	FLOORSMIN_MODE	FLOORSMAX_AVG	0.722966
1602	LIVINGAREA_AVG	BASEMENTAREA_AVG	0.695154
2710	BASEMENTAREA_MEDI	LIVINGAREA_AVG	0.695129
3338	LIVINGAREA_MEDI	BASEMENTAREA_MEDI	0.694073
2470	LIVINGAREA_MODE	BASEMENTAREA_MODE	0.693588
3310	LIVINGAREA_MEDI	BASEMENTAREA_AVG	0.691838
2724	BASEMENTAREA_MEDI	LIVINGAREA_MODE	0.683185
2727	BASEMENTAREA_MEDI	APARTMENTS_MEDI	0.682731
991	BASEMENTAREA_AVG	APARTMENTS_AVG	0.681712
2699	BASEMENTAREA_MEDI	APARTMENTS_AVG	0.681551
1859	BASEMENTAREA_MODE	APARTMENTS_MODE	0.681342
2639	APARTMENTS_MEDI	BASEMENTAREA_AVG	0.680320
1362	FLOORSMAX_AVG	ELEVATORS_AVG	0.680080
2456	LIVINGAREA_MODE	BASEMENTAREA_AVG	0.679499
3070	FLOORSMAX_MEDI	ELEVATORS_AVG	0.677748
1856	BASEMENTAREA_MODE	LIVINGAREA_AVG	0.676864
2950	ELEVATORS_MEDI	FLOORSMAX_AVG	0.676434
3324	LIVINGAREA_MEDI	BASEMENTAREA_MODE	0.676385
3098	FLOORSMAX_MEDI	ELEVATORS_MEDI	0.675868
3493	TOTALAREA_MODE	BASEMENTAREA_AVG	0.673406
2216	FLOORSMAX_MODE	ELEVATORS_AVG	0.670997
3521	TOTALAREA_MODE	BASEMENTAREA_MEDI	0.670901
2713	BASEMENTAREA_MEDI	APARTMENTS_MODE	0.670827
2964	ELEVATORS_MEDI	FLOORSMAX_MODE	0.669300
1785	APARTMENTS_MODE	BASEMENTAREA_AVG	0.667797
2653	APARTMENTS_MEDI	BASEMENTAREA_MODE	0.666497
1845	BASEMENTAREA_MODE	APARTMENTS_AVG	0.664205
2230	FLOORSMAX_MODE	ELEVATORS_MODE	0.661472
2409	LIVINGAPARTMENTS_MODE	BASEMENTAREA_MODE	0.657269
2165	ENTRANCES_MODE	BASEMENTAREA_MODE	0.656745
2096	ELEVATORS_MODE	FLOORSMAX_AVG	0.656743
3084	FLOORSMAX_MEDI	ELEVATORS_MODE	0.656085
1851	BASEMENTAREA_MODE	ENTRANCES_AVG	0.655926
2705	BASEMENTAREA_MEDI	ENTRANCES_AVG	0.655063
3019	ENTRANCES_MEDI	BASEMENTAREA_MODE	0.654844
2723	BASEMENTAREA_MEDI	LIVINGAPARTMENTS_MODE	0.653997
3277	LIVINGAPARTMENTS_MEDI	BASEMENTAREA_MEDI	0.653879
3033	ENTRANCES_MEDI	BASEMENTAREA_MEDI	0.653525
1297	ENTRANCES_AVG	BASEMENTAREA_AVG	0.652845
3249	LIVINGAPARTMENTS_MEDI	BASEMENTAREA_AVG	0.651098
3507	TOTALAREA_MODE	BASEMENTAREA_MODE	0.650905
2709	BASEMENTAREA_MEDI	LIVINGAPARTMENTS_AVG	0.649699
2395	LIVINGAPARTMENTS_MODE	BASEMENTAREA_AVG	0.649393
1541	LIVINGAPARTMENTS_AVG	BASEMENTAREA_AVG	0.649292
3005	ENTRANCES_MEDI	BASEMENTAREA_AVG	0.648754
2719	BASEMENTAREA_MEDI	ENTRANCES_MODE	0.633048
3263	LIVINGAPARTMENTS_MEDI	BASEMENTAREA_MODE	0.632473
3499	TOTALAREA_MODE	FLOORSMAX_AVG	0.632121
1608	LIVINGAREA_AVG	FLOORSMAX_AVG	0.631623
3076	FLOORSMAX_MEDI	LIVINGAREA_AVG	0.629591
3527	TOTALAREA_MODE	FLOORSMAX_MEDI	0.629423
3316	LIVINGAREA_MEDI	FLOORSMAX_AVG	0.628424
1855	BASEMENTAREA_MODE	LIVINGAPARTMENTS_AVG	0.628040
2151	ENTRANCES_MODE	BASEMENTAREA_AVG	0.627817
3344	LIVINGAREA_MEDI	FLOORSMAX_MEDI	0.627624
2222	FLOORSMAX_MODE	LIVINGAREA_AVG	0.627241
3330	LIVINGAREA_MEDI	FLOORSMAX_MODE	0.625625
3513	TOTALAREA_MODE	FLOORSMAX_MODE	0.624915
2475	LIVINGAREA_MODE	ENTRANCES_MODE	0.621397
2461	LIVINGAREA_MODE	ENTRANCES_AVG	0.621177
3029	ENTRANCES_MEDI	LIVINGAREA_MODE	0.620609
1357	FLOORSMAX_AVG	APARTMENTS_AVG	0.618742
3315	LIVINGAREA_MEDI	ENTRANCES_AVG	0.618092
3343	LIVINGAREA_MEDI	ENTRANCES_MEDI	0.617806
1607	LIVINGAREA_AVG	ENTRANCES_AVG	0.617606
3065	FLOORSMAX_MEDI	APARTMENTS_AVG	0.616645
433	DAYS_EMPLOYED	DAYS_BIRTH	0.615939
2645	APARTMENTS_MEDI	FLOORSMAX_AVG	0.615532
2164	ENTRANCES_MODE	APARTMENTS_MODE	0.614903
3093	FLOORSMAX_MEDI	APARTMENTS_MEDI	0.614634
2211	FLOORSMAX_MODE	APARTMENTS_AVG	0.614561
3015	ENTRANCES_MEDI	LIVINGAREA_AVG	0.613990
2659	APARTMENTS_MEDI	FLOORSMAX_MODE	0.612911
3018	ENTRANCES_MEDI	APARTMENTS_MODE	0.611534
1790	APARTMENTS_MODE	ENTRANCES_AVG	0.611277
2644	APARTMENTS_MEDI	ENTRANCES_AVG	0.610779
1296	ENTRANCES_AVG	APARTMENTS_AVG	0.610692
3032	ENTRANCES_MEDI	APARTMENTS_MEDI	0.610665
2476	LIVINGAREA_MODE	FLOORSMAX_MODE	0.607347
3004	ENTRANCES_MEDI	APARTMENTS_AVG	0.607231
738	EXT_SOURCE_1	DAYS_BIRTH	0.600492
2462	LIVINGAREA_MODE	FLOORSMAX_AVG	0.598360
3090	FLOORSMAX_MEDI	LIVINGAREA_MODE	0.597595
2225	FLOORSMAX_MODE	APARTMENTS_MODE	0.596504
3498	TOTALAREA_MODE	ENTRANCES_AVG	0.593474
3329	LIVINGAREA_MEDI	ENTRANCES_MODE	0.593359
1547	LIVINGAPARTMENTS_AVG	FLOORSMAX_AVG	0.589843
3255	LIVINGAPARTMENTS_MEDI	FLOORSMAX_AVG	0.589165
2161	ENTRANCES_MODE	LIVINGAREA_AVG	0.588945
3526	TOTALAREA_MODE	ENTRANCES_MEDI	0.587157
2658	APARTMENTS_MEDI	ENTRANCES_MODE	0.586964
3283	LIVINGAPARTMENTS_MEDI	FLOORSMAX_MEDI	0.586745
3075	FLOORSMAX_MEDI	LIVINGAPARTMENTS_AVG	0.586583
1791	APARTMENTS_MODE	FLOORSMAX_AVG	0.586504
3079	FLOORSMAX_MEDI	APARTMENTS_MODE	0.585628
3269	LIVINGAPARTMENTS_MEDI	FLOORSMAX_MODE	0.583228
2221	FLOORSMAX_MODE	LIVINGAPARTMENTS_AVG	0.582849
2150	ENTRANCES_MODE	APARTMENTS_AVG	0.582479
2400	LIVINGAPARTMENTS_MODE	ENTRANCES_AVG	0.573465
2415	LIVINGAPARTMENTS_MODE	FLOORSMAX_MODE	0.572674
3028	ENTRANCES_MEDI	LIVINGAPARTMENTS_MODE	0.572349
2401	LIVINGAPARTMENTS_MODE	FLOORSMAX_AVG	0.568441
3089	FLOORSMAX_MEDI	LIVINGAPARTMENTS_MODE	0.566405
3254	LIVINGAPARTMENTS_MEDI	ENTRANCES_AVG	0.566080
2414	LIVINGAPARTMENTS_MODE	ENTRANCES_MODE	0.565401
1236	ELEVATORS_AVG	BASEMENTAREA_AVG	0.564902
3282	LIVINGAPARTMENTS_MEDI	ENTRANCES_MEDI	0.564775
1546	LIVINGAPARTMENTS_AVG	ENTRANCES_AVG	0.563805
2972	ELEVATORS_MEDI	BASEMENTAREA_MEDI	0.563126
2704	BASEMENTAREA_MEDI	ELEVATORS_AVG	0.563083
2944	ELEVATORS_MEDI	BASEMENTAREA_AVG	0.562466
3014	ENTRANCES_MEDI	LIVINGAPARTMENTS_AVG	0.559816
3512	TOTALAREA_MODE	ENTRANCES_MODE	0.559791
2718	BASEMENTAREA_MEDI	ELEVATORS_MODE	0.556170
2090	ELEVATORS_MODE	BASEMENTAREA_AVG	0.555015
2104	ELEVATORS_MODE	BASEMENTAREA_MODE	0.554090
3496	TOTALAREA_MODE	COMMONAREA_AVG	0.553295
3524	TOTALAREA_MODE	COMMONAREA_MEDI	0.552846
3341	LIVINGAREA_MEDI	COMMONAREA_MEDI	0.550937
2893	COMMONAREA_MEDI	LIVINGAREA_AVG	0.550281

Conclusion :

Si dos variables están fuertemente correlacionadas, esto puede indicar multicolinealidad, donde las variables predicen esencialmente la misma información.

Aqui vemos en orden las variables mas correladas entre si, se nota que en la mayoria de los casos son variables de 'tipo de vivienda' que aportan detalles sobre el hogar, lo que tiene sentido porque son variables relacionadas entre si y muy detalladas, que mas o menos describen con detalle la vivienda.

• Otras variables relacionadas son AMT_ANNUITY y AMT_CREDIT, y AMT_ANNUITY y AMT_GOODS_PRICE que van relacionando por proporcionalidad, porque AMT_ANNUITY es proporcional a AMT_CREDIT, ya que una cuota mensual más alta está típicamente asociada a un monto de crédito más alto.

Se podria, segun del modelo, considerar removar algunas, o juntarles en una nueva variable para reducir el numero de variables y la multicolinealidad. Se puede tambien considerar usar el PCA...

MISSING VALUES

eda.get_percent_null_values_target(data_train, continuous_vars, target='TARGET')

	Category_0	variable	sum_null_values	porcentaje_sum_null_values	Category_1
0	1.000000	AMT_ANNUITY	11	0.000045	NaN
1	0.919431	AMT_GOODS_PRICE	211	0.000858	0.080569
2	0.914944	OWN_CAR_AGE	162329	0.659853	0.085056
3	0.914595	EXT_SOURCE_1	138528	0.563104	0.085405
4	0.914729	EXT_SOURCE_2	516	0.002097	0.085271
5	0.907155	EXT_SOURCE_3	48974	0.199075	0.092845
6	0.908756	APARTMENTS_AVG	124764	0.507154	0.091244
7	0.911016	BASEMENTAREA_AVG	143880	0.584859	0.088984
8	0.908192	YEARS_BEGINEXPLUATATION_AVG	119935	0.487525	0.091808
9	0.913332	YEARS_BUILD_AVG	163544	0.664791	0.086668
10	0.914398	COMMONAREA_AVG	171830	0.698473	0.085602
11	0.909222	ELEVATORS_AVG	131001	0.532507	0.090778
12	0.908444	ENTRANCES_AVG	123739	0.502988	0.091556
13	0.908375	FLOORSMAX_AVG	122325	0.497240	0.091625
14	0.913850	FLOORSMIN_AVG	166871	0.678315	0.086150
15	0.911949	LANDAREA_AVG	146029	0.593595	0.088051
16	0.913899	LIVINGAPARTMENTS_AVG	168151	0.683518	0.086101
17	0.908792	LIVINGAREA_AVG	123356	0.501431	0.091208
18	0.914308	NONLIVINGAPARTMENTS_AVG	170798	0.694278	0.085692
19	0.909864	NONLIVINGAREA_AVG	135617	0.551271	0.090136
20	0.908756	APARTMENTS_MODE	124764	0.507154	0.091244
21	0.911016	BASEMENTAREA_MODE	143880	0.584859	0.088984
22	0.908192	YEARS_BEGINEXPLUATATION_MODE	119935	0.487525	0.091808
23	0.913332	YEARS_BUILD_MODE	163544	0.664791	0.086668
24	0.914398	COMMONAREA_MODE	171830	0.698473	0.085602
25	0.909222	ELEVATORS_MODE	131001	0.532507	0.090778
26	0.908444	ENTRANCES_MODE	123739	0.502988	0.091556
27	0.908375	FLOORSMAX_MODE	122325	0.497240	0.091625
28	0.913850	FLOORSMIN_MODE	166871	0.678315	0.086150
29	0.911949	LANDAREA_MODE	146029	0.593595	0.088051
30	0.913899	LIVINGAPARTMENTS_MODE	168151	0.683518	0.086101
31	0.908792	LIVINGAREA_MODE	123356	0.501431	0.091208
32	0.914308	NONLIVINGAPARTMENTS_MODE	170798	0.694278	0.085692
33	0.909864	NONLIVINGAREA_MODE	135617	0.551271	0.090136
34	0.908756	APARTMENTS_MEDI	124764	0.507154	0.091244
35	0.911016	BASEMENTAREA_MEDI	143880	0.584859	0.088984
36	0.908192	YEARS_BEGINEXPLUATATION_MEDI	119935	0.487525	0.091808
37	0.913332	YEARS_BUILD_MEDI	163544	0.664791	0.086668
38	0.914398	COMMONAREA_MEDI	171830	0.698473	0.085602
39	0.909222	ELEVATORS_MEDI	131001	0.532507	0.090778
40	0.908444	ENTRANCES_MEDI	123739	0.502988	0.091556
41	0.908375	FLOORSMAX_MEDI	122325	0.497240	0.091625
42	0.913850	FLOORSMIN_MEDI	166871	0.678315	0.086150
43	0.911949	LANDAREA_MEDI	146029	0.593595	0.088051
44	0.913899	LIVINGAPARTMENTS_MEDI	168151	0.683518	0.086101
45	0.908792	LIVINGAREA_MEDI	123356	0.501431	0.091208
46	0.914308	NONLIVINGAPARTMENTS_MEDI	170798	0.694278	0.085692
47	0.909864	NONLIVINGAREA_MEDI	135617	0.551271	0.090136
48	0.907951	TOTALAREA_MODE	118643	0.482273	0.092049
49	0.896709	AMT_REQ_CREDIT_BUREAU_MON	33391	0.135731	0.103291
50	0.896709	AMT_REQ_CREDIT_BUREAU_YEAR	33391	0.135731	0.103291

Si una variable importante tiene muchos valores faltantes, puede significar que el cliente no entregó un documento necesario, lo que podría ser senal de mayor riesgo. También podría ser que esa información no era obligatoria al solicitar el préstamo. En cualquier caso, pueden dar pistas sobre el comportamiento o riesgo del cliente.

QUE ESTRATREGIA DECIDI SEGUIR ?

En algunos campos se debe elegir particularmente

Decidimos agrupar variables de vivienda, y variables mas especificas para imputar de dos maneras diferentes:

Utilizamos dos métodos para tratar los valores faltantes según las características de los datos.

• La imputación con la media la aplico a las variables relacionadas con la vivienda y la descripción de los hogares, ya que estas variables son relativamente homogéneas y independientes de otros factores, lo que hace que este método sea simple y efectivo.

• Por otro lado, usé el SimpleImputer con la strategy de mediana, para las columnas continuas que quedaban como las de AMT_ANNUITY, OWN_CAR_AGE y EXT_SOURCE_1...

Ya que es una técnica robusta con valores atípicos, que son mas o menos comunes en estas variables. La mediana garantiza una imputación mas rápida y razonable, evitando distorsiones debidas a valores extremos.

Además, permite un flujo consistente y escalable entre conjuntos de entrenamiento y prueba, asegurando que las mismas imputaciones se apliquen de forma uniforme. Si se necesita mayor precisión o se identifican relaciones complejas entre variables, podría implementarse un KNearestNeighbors para predecir mejor valores missing con sus dependencias asociadas a sus 'vecinos'os.

# copio los datos originales
data_train_input = data_train.copy()
data_test_input = data_test.copy()

#alisto mis columnas continuas relacionadas con la vivienda cuales nulos cambiaré con la media:
housing_continues_columns = [
    'APARTMENTS_AVG', 'BASEMENTAREA_AVG', 'YEARS_BEGINEXPLUATATION_AVG', 'YEARS_BUILD_AVG', 
    'COMMONAREA_AVG', 'ELEVATORS_AVG', 'ENTRANCES_AVG', 'FLOORSMAX_AVG', 'FLOORSMIN_AVG', 
    'LANDAREA_AVG', 'LIVINGAPARTMENTS_AVG', 'LIVINGAREA_AVG', 'NONLIVINGAPARTMENTS_AVG', 
    'NONLIVINGAREA_AVG', 'APARTMENTS_MODE', 'BASEMENTAREA_MODE', 'YEARS_BEGINEXPLUATATION_MODE', 
    'YEARS_BUILD_MODE', 'COMMONAREA_MODE', 'ELEVATORS_MODE', 'ENTRANCES_MODE', 
    'FLOORSMAX_MODE', 'FLOORSMIN_MODE', 'LANDAREA_MODE', 'LIVINGAPARTMENTS_MODE', 
    'LIVINGAREA_MODE', 'NONLIVINGAPARTMENTS_MODE', 'NONLIVINGAREA_MODE', 'APARTMENTS_MEDI', 
    'BASEMENTAREA_MEDI', 'YEARS_BEGINEXPLUATATION_MEDI', 'YEARS_BUILD_MEDI', 'COMMONAREA_MEDI', 
    'ELEVATORS_MEDI', 'ENTRANCES_MEDI', 'FLOORSMAX_MEDI', 'FLOORSMIN_MEDI', 'LANDAREA_MEDI', 
    'LIVINGAPARTMENTS_MEDI', 'LIVINGAREA_MEDI', 'NONLIVINGAPARTMENTS_MEDI', 
    'NONLIVINGAREA_MEDI', 'TOTALAREA_MODE'
]

#IMPUTANDO CON LA MEDIA EN LOS DOS CONJUNTOS DE MI DATA:
data_train_input[housing_continues_columns] = data_train[housing_continues_columns].fillna(data_train[housing_continues_columns].mean())
data_test_input[housing_continues_columns] = data_test[housing_continues_columns].fillna(data_test[housing_continues_columns].mean())

#Otras columnas que pondré con una imputacion con mediana:
remaining_continues_columns = [
    'AMT_ANNUITY', 'AMT_GOODS_PRICE', 'OWN_CAR_AGE', 'EXT_SOURCE_1', 
    'EXT_SOURCE_2', 'EXT_SOURCE_3', 'AMT_REQ_CREDIT_BUREAU_MON', 'AMT_REQ_CREDIT_BUREAU_YEAR'
]

#escalado de los datos (bloques para mas eficiencia)
scaler = StandardScaler()

def scale_and_impute(data, columns, scaler, imputer):
    scaled_data = scaler.fit_transform(data[columns]) 
    return imputer.fit_transform(scaled_data)          

#USO EL METODO DE LA MEDIANA:
simple_imputer = SimpleImputer(strategy="median")

#imputar con escalado para mas facilidad
data_train_imputed = scale_and_impute(data_train_input, remaining_continues_columns, scaler, simple_imputer)
data_test_imputed = scale_and_impute(data_test_input, remaining_continues_columns, scaler, simple_imputer)

#dataframe completos
data_train_imputed_df = pd.DataFrame(data_train_imputed, columns=remaining_continues_columns, index=data_train_input.index)
data_test_imputed_df = pd.DataFrame(data_test_imputed, columns=remaining_continues_columns, index=data_test_input.index)

#asignar nuevos datos al dataframe que he copiado en primer lugar
data_train_input[remaining_continues_columns] = data_train_imputed_df
data_test_input[remaining_continues_columns] = data_test_imputed_df

#COMPROBACION

print("\nValores nulos después de la imputación en train (remaining_continues_columns):")
print(data_train_input[remaining_continues_columns].isnull().sum())

print("\nValores nulos después de la imputación en test (remaining_continues_columns):")
print(data_test_input[remaining_continues_columns].isnull().sum())

Valores nulos después de la imputación en train (remaining_continues_columns):
AMT_ANNUITY                   0
AMT_GOODS_PRICE               0
OWN_CAR_AGE                   0
EXT_SOURCE_1                  0
EXT_SOURCE_2                  0
EXT_SOURCE_3                  0
AMT_REQ_CREDIT_BUREAU_MON     0
AMT_REQ_CREDIT_BUREAU_YEAR    0
dtype: int64

Valores nulos después de la imputación en test (remaining_continues_columns):
AMT_ANNUITY                   0
AMT_GOODS_PRICE               0
OWN_CAR_AGE                   0
EXT_SOURCE_1                  0
EXT_SOURCE_2                  0
EXT_SOURCE_3                  0
AMT_REQ_CREDIT_BUREAU_MON     0
AMT_REQ_CREDIT_BUREAU_YEAR    0
dtype: int64

Variables categoricas

print("Cantidad de valores nulos en 'TARGET':", data_train['TARGET'].isnull().sum())

Cantidad de valores nulos en 'TARGET': 0

categorical_vars

['NAME_CONTRACT_TYPE',
 'CODE_GENDER',
 'FLAG_OWN_CAR',
 'FLAG_OWN_REALTY',
 'CNT_CHILDREN',
 'NAME_TYPE_SUITE',
 'NAME_INCOME_TYPE',
 'NAME_EDUCATION_TYPE',
 'NAME_FAMILY_STATUS',
 'NAME_HOUSING_TYPE',
 'FLAG_MOBIL',
 'FLAG_EMP_PHONE',
 'FLAG_WORK_PHONE',
 'FLAG_CONT_MOBILE',
 'FLAG_PHONE',
 'FLAG_EMAIL',
 'OCCUPATION_TYPE',
 'CNT_FAM_MEMBERS',
 'REGION_RATING_CLIENT',
 'REGION_RATING_CLIENT_W_CITY',
 'WEEKDAY_APPR_PROCESS_START',
 'REG_REGION_NOT_LIVE_REGION',
 'REG_REGION_NOT_WORK_REGION',
 'LIVE_REGION_NOT_WORK_REGION',
 'REG_CITY_NOT_LIVE_CITY',
 'REG_CITY_NOT_WORK_CITY',
 'LIVE_CITY_NOT_WORK_CITY',
 'ORGANIZATION_TYPE',
 'FONDKAPREMONT_MODE',
 'HOUSETYPE_MODE',
 'WALLSMATERIAL_MODE',
 'EMERGENCYSTATE_MODE',
 'FLAG_DOCUMENT_2',
 'FLAG_DOCUMENT_3',
 'FLAG_DOCUMENT_4',
 'FLAG_DOCUMENT_5',
 'FLAG_DOCUMENT_6',
 'FLAG_DOCUMENT_7',
 'FLAG_DOCUMENT_8',
 'FLAG_DOCUMENT_9',
 'FLAG_DOCUMENT_10',
 'FLAG_DOCUMENT_11',
 'FLAG_DOCUMENT_12',
 'FLAG_DOCUMENT_13',
 'FLAG_DOCUMENT_14',
 'FLAG_DOCUMENT_15',
 'FLAG_DOCUMENT_16',
 'FLAG_DOCUMENT_17',
 'FLAG_DOCUMENT_18',
 'FLAG_DOCUMENT_19',
 'FLAG_DOCUMENT_20',
 'FLAG_DOCUMENT_21',
 'AMT_REQ_CREDIT_BUREAU_HOUR',
 'AMT_REQ_CREDIT_BUREAU_DAY',
 'AMT_REQ_CREDIT_BUREAU_WEEK',
 'AMT_REQ_CREDIT_BUREAU_QRT',
 'TARGET']

#conteos antes de tratar los datos categoricos :

print("missing de categoricas antes de transformar:")
print(data_train_input[categorical_vars].isnull().sum())


print("\nTipo de cada variable:")
print(data_train_input[categorical_vars].dtypes)


print("\nValores unicos de cada variable:")
print(data_train_input[categorical_vars].nunique())

missing de categoricas antes de transformar:
NAME_CONTRACT_TYPE                  0
CODE_GENDER                         3
FLAG_OWN_CAR                        0
FLAG_OWN_REALTY                     0
CNT_CHILDREN                        0
NAME_TYPE_SUITE                  1003
NAME_INCOME_TYPE                    0
NAME_EDUCATION_TYPE                 0
NAME_FAMILY_STATUS                  0
NAME_HOUSING_TYPE                   0
FLAG_MOBIL                          0
FLAG_EMP_PHONE                      0
FLAG_WORK_PHONE                     0
FLAG_CONT_MOBILE                    0
FLAG_PHONE                          0
FLAG_EMAIL                          0
OCCUPATION_TYPE                 76942
CNT_FAM_MEMBERS                     2
REGION_RATING_CLIENT                0
REGION_RATING_CLIENT_W_CITY         0
WEEKDAY_APPR_PROCESS_START          0
REG_REGION_NOT_LIVE_REGION          0
REG_REGION_NOT_WORK_REGION          0
LIVE_REGION_NOT_WORK_REGION         0
REG_CITY_NOT_LIVE_CITY              0
REG_CITY_NOT_WORK_CITY              0
LIVE_CITY_NOT_WORK_CITY             0
ORGANIZATION_TYPE                   0
FONDKAPREMONT_MODE             168191
HOUSETYPE_MODE                 123367
WALLSMATERIAL_MODE             125005
EMERGENCYSTATE_MODE            116513
FLAG_DOCUMENT_2                     0
FLAG_DOCUMENT_3                     0
FLAG_DOCUMENT_4                     0
FLAG_DOCUMENT_5                     0
FLAG_DOCUMENT_6                     0
FLAG_DOCUMENT_7                     0
FLAG_DOCUMENT_8                     0
FLAG_DOCUMENT_9                     0
FLAG_DOCUMENT_10                    0
FLAG_DOCUMENT_11                    0
FLAG_DOCUMENT_12                    0
FLAG_DOCUMENT_13                    0
FLAG_DOCUMENT_14                    0
FLAG_DOCUMENT_15                    0
FLAG_DOCUMENT_16                    0
FLAG_DOCUMENT_17                    0
FLAG_DOCUMENT_18                    0
FLAG_DOCUMENT_19                    0
FLAG_DOCUMENT_20                    0
FLAG_DOCUMENT_21                    0
AMT_REQ_CREDIT_BUREAU_HOUR      33391
AMT_REQ_CREDIT_BUREAU_DAY       33391
AMT_REQ_CREDIT_BUREAU_WEEK      33391
AMT_REQ_CREDIT_BUREAU_QRT       33391
TARGET                              0
dtype: int64

Tipo de cada variable:
NAME_CONTRACT_TYPE              object
CODE_GENDER                     object
FLAG_OWN_CAR                    object
FLAG_OWN_REALTY                 object
CNT_CHILDREN                     int64
NAME_TYPE_SUITE                 object
NAME_INCOME_TYPE                object
NAME_EDUCATION_TYPE             object
NAME_FAMILY_STATUS              object
NAME_HOUSING_TYPE               object
FLAG_MOBIL                       int64
FLAG_EMP_PHONE                   int64
FLAG_WORK_PHONE                  int64
FLAG_CONT_MOBILE                 int64
FLAG_PHONE                       int64
FLAG_EMAIL                       int64
OCCUPATION_TYPE                 object
CNT_FAM_MEMBERS                float64
REGION_RATING_CLIENT             int64
REGION_RATING_CLIENT_W_CITY      int64
WEEKDAY_APPR_PROCESS_START      object
REG_REGION_NOT_LIVE_REGION       int64
REG_REGION_NOT_WORK_REGION       int64
LIVE_REGION_NOT_WORK_REGION      int64
REG_CITY_NOT_LIVE_CITY           int64
REG_CITY_NOT_WORK_CITY           int64
LIVE_CITY_NOT_WORK_CITY          int64
ORGANIZATION_TYPE               object
FONDKAPREMONT_MODE              object
HOUSETYPE_MODE                  object
WALLSMATERIAL_MODE              object
EMERGENCYSTATE_MODE             object
FLAG_DOCUMENT_2                  int64
FLAG_DOCUMENT_3                  int64
FLAG_DOCUMENT_4                  int64
FLAG_DOCUMENT_5                  int64
FLAG_DOCUMENT_6                  int64
FLAG_DOCUMENT_7                  int64
FLAG_DOCUMENT_8                  int64
FLAG_DOCUMENT_9                  int64
FLAG_DOCUMENT_10                 int64
FLAG_DOCUMENT_11                 int64
FLAG_DOCUMENT_12                 int64
FLAG_DOCUMENT_13                 int64
FLAG_DOCUMENT_14                 int64
FLAG_DOCUMENT_15                 int64
FLAG_DOCUMENT_16                 int64
FLAG_DOCUMENT_17                 int64
FLAG_DOCUMENT_18                 int64
FLAG_DOCUMENT_19                 int64
FLAG_DOCUMENT_20                 int64
FLAG_DOCUMENT_21                 int64
AMT_REQ_CREDIT_BUREAU_HOUR     float64
AMT_REQ_CREDIT_BUREAU_DAY      float64
AMT_REQ_CREDIT_BUREAU_WEEK     float64
AMT_REQ_CREDIT_BUREAU_QRT      float64
TARGET                           int64
dtype: object

Valores unicos de cada variable:
NAME_CONTRACT_TYPE              2
CODE_GENDER                     2
FLAG_OWN_CAR                    2
FLAG_OWN_REALTY                 2
CNT_CHILDREN                   15
NAME_TYPE_SUITE                 7
NAME_INCOME_TYPE                8
NAME_EDUCATION_TYPE             5
NAME_FAMILY_STATUS              6
NAME_HOUSING_TYPE               6
FLAG_MOBIL                      1
FLAG_EMP_PHONE                  2
FLAG_WORK_PHONE                 2
FLAG_CONT_MOBILE                2
FLAG_PHONE                      2
FLAG_EMAIL                      2
OCCUPATION_TYPE                18
CNT_FAM_MEMBERS                17
REGION_RATING_CLIENT            3
REGION_RATING_CLIENT_W_CITY     3
WEEKDAY_APPR_PROCESS_START      7
REG_REGION_NOT_LIVE_REGION      2
REG_REGION_NOT_WORK_REGION      2
LIVE_REGION_NOT_WORK_REGION     2
REG_CITY_NOT_LIVE_CITY          2
REG_CITY_NOT_WORK_CITY          2
LIVE_CITY_NOT_WORK_CITY         2
ORGANIZATION_TYPE              58
FONDKAPREMONT_MODE              4
HOUSETYPE_MODE                  3
WALLSMATERIAL_MODE              7
EMERGENCYSTATE_MODE             2
FLAG_DOCUMENT_2                 2
FLAG_DOCUMENT_3                 2
FLAG_DOCUMENT_4                 2
FLAG_DOCUMENT_5                 2
FLAG_DOCUMENT_6                 2
FLAG_DOCUMENT_7                 2
FLAG_DOCUMENT_8                 2
FLAG_DOCUMENT_9                 2
FLAG_DOCUMENT_10                2
FLAG_DOCUMENT_11                2
FLAG_DOCUMENT_12                2
FLAG_DOCUMENT_13                2
FLAG_DOCUMENT_14                2
FLAG_DOCUMENT_15                2
FLAG_DOCUMENT_16                2
FLAG_DOCUMENT_17                2
FLAG_DOCUMENT_18                2
FLAG_DOCUMENT_19                2
FLAG_DOCUMENT_20                2
FLAG_DOCUMENT_21                2
AMT_REQ_CREDIT_BUREAU_HOUR      5
AMT_REQ_CREDIT_BUREAU_DAY       9
AMT_REQ_CREDIT_BUREAU_WEEK      9
AMT_REQ_CREDIT_BUREAU_QRT      10
TARGET                          2
dtype: int64

#bucle para tener las correlaciones categoricas de cada variable categorica respeto a la variable objetivo:
for var in categorical_vars:
    print('------------------------------------------------------------')
    print(f"Confusion matrix parfa {var} con respeto a TARGET:")
    confusion_matrix = pd.crosstab(data_train['TARGET'], data_train_input[var])
    print(confusion_matrix)
    #coeficiente de Cramers v:
    cramers_v_value = eda.cramers_v(confusion_matrix.values)
    print(f"Cramér's V for {var}: {cramers_v_value}\n")

------------------------------------------------------------
Confusion matrix parfa NAME_CONTRACT_TYPE con respeto a TARGET:
NAME_CONTRACT_TYPE  CASH LOANS  REVOLVING LOANS
TARGET                                         
0                       204106            22042
1                        18590             1270
Cramér's V for NAME_CONTRACT_TYPE: 0.0310865205023682

------------------------------------------------------------
Confusion matrix parfa CODE_GENDER con respeto a TARGET:
CODE_GENDER  FEMALE   MALE
TARGET                    
0            150573  75572
1             11307   8553
Cramér's V for CODE_GENDER: 0.05535981073628126

------------------------------------------------------------
Confusion matrix parfa FLAG_OWN_CAR con respeto a TARGET:
FLAG_OWN_CAR      NO    YES
TARGET                     
0             148517  77631
1              13807   6053
Cramér's V for FLAG_OWN_CAR: 0.022025230924775974

------------------------------------------------------------
Confusion matrix parfa FLAG_OWN_REALTY con respeto a TARGET:
FLAG_OWN_REALTY     NO     YES
TARGET                        
0                69132  157016
1                 6341   13519
Cramér's V for FLAG_OWN_REALTY: 0.007754761693103105

------------------------------------------------------------
Confusion matrix parfa CNT_CHILDREN con respeto a TARGET:
CNT_CHILDREN      0      1      2     3    4   5   6   7   8   9   10  11  12  \
TARGET                                                                          
0             159018  44541  19528  2674  302  60  11   5   2   0   1   0   2   
1              13325   4327   1861   287   46   6   5   0   0   2   0   1   0   

CNT_CHILDREN  14  19  
TARGET                
0              2   2  
1              0   0  
Cramér's V for CNT_CHILDREN: 0.023388575925143276

------------------------------------------------------------
Confusion matrix parfa NAME_TYPE_SUITE con respeto a TARGET:
NAME_TYPE_SUITE  CHILDREN  FAMILY  GROUP OF PEOPLE  OTHER_A  OTHER_B  \
TARGET                                                                 
0                    2468   29724              200      637     1266   
1                     193    2399               17       62      140   

NAME_TYPE_SUITE  SPOUSE, PARTNER  UNACCOMPANIED  
TARGET                                           
0                           8286         182618  
1                            713          16282  
Cramér's V for NAME_TYPE_SUITE: 0.009732772442277339

------------------------------------------------------------
Confusion matrix parfa NAME_INCOME_TYPE con respeto a TARGET:
NAME_INCOME_TYPE  BUSINESSMAN  COMMERCIAL ASSOCIATE  MATERNITY LEAVE  \
TARGET                                                                 
0                          10                 53077                3   
1                           0                  4324                1   

NAME_INCOME_TYPE  PENSIONER  STATE SERVANT  STUDENT  UNEMPLOYED  WORKING  
TARGET                                                                    
0                     41765          16381       13          13   114886  
1                      2367           1025        0           5    12138  
Cramér's V for NAME_INCOME_TYPE: 0.062250216935764706

------------------------------------------------------------
Confusion matrix parfa NAME_EDUCATION_TYPE con respeto a TARGET:
NAME_EDUCATION_TYPE  ACADEMIC DEGREE  HIGHER EDUCATION  INCOMPLETE HIGHER  \
TARGET                                                                      
0                                127             56774               7539   
1                                  2              3192                695   

NAME_EDUCATION_TYPE  LOWER SECONDARY  SECONDARY / SECONDARY SPECIAL  
TARGET                                                               
0                               2665                         159043  
1                                338                          15633  
Cramér's V for NAME_EDUCATION_TYPE: 0.058359656764498065

------------------------------------------------------------
Confusion matrix parfa NAME_FAMILY_STATUS con respeto a TARGET:
NAME_FAMILY_STATUS  CIVIL MARRIAGE  MARRIED  SEPARATED  SINGLE / NOT MARRIED  \
TARGET                                                                         
0                            21364   145299      14534                 32785   
1                             2375    11842       1329                  3557   

NAME_FAMILY_STATUS  UNKNOWN  WIDOW  
TARGET                              
0                         2  12164  
1                         0    757  
Cramér's V for NAME_FAMILY_STATUS: 0.04067121943031738

------------------------------------------------------------
Confusion matrix parfa NAME_HOUSING_TYPE con respeto a TARGET:
NAME_HOUSING_TYPE  CO-OP APARTMENT  HOUSE / APARTMENT  MUNICIPAL APARTMENT  \
TARGET                                                                       
0                              834             201252                 8228   
1                               72              17008                  753   

NAME_HOUSING_TYPE  OFFICE APARTMENT  RENTED APARTMENT  WITH PARENTS  
TARGET                                                               
0                              1950              3460         10424  
1                               137               484          1406  
Cramér's V for NAME_HOUSING_TYPE: 0.03777646043528446

------------------------------------------------------------
Confusion matrix parfa FLAG_MOBIL con respeto a TARGET:
FLAG_MOBIL       1
TARGET            
0           226148
1            19860
Cramér's V for FLAG_MOBIL: nan

------------------------------------------------------------
Confusion matrix parfa FLAG_EMP_PHONE con respeto a TARGET:
FLAG_EMP_PHONE      0       1
TARGET                       
0               41778  184370
1                2373   17487
Cramér's V for FLAG_EMP_PHONE: 0.04625807891255162

------------------------------------------------------------
Confusion matrix parfa FLAG_WORK_PHONE con respeto a TARGET:

C:\Users\ninao\Desktop\ML_practica1EDA\notebooks\../src\eda_utils.py:549: RuntimeWarning: invalid value encountered in double_scalars
  return np.sqrt(phi2corr / min((kcorr-1), (rcorr-1)))

FLAG_WORK_PHONE       0      1
TARGET                        
0                181826  44322
1                 15095   4765
Cramér's V for FLAG_WORK_PHONE: 0.02986670896835815

------------------------------------------------------------
Confusion matrix parfa FLAG_CONT_MOBILE con respeto a TARGET:
FLAG_CONT_MOBILE    0       1
TARGET                       
0                 427  225721
1                  36   19824
Cramér's V for FLAG_CONT_MOBILE: 0.0

------------------------------------------------------------
Confusion matrix parfa FLAG_PHONE con respeto a TARGET:
FLAG_PHONE       0      1
TARGET                   
0           161863  64285
1            14939   4921
Cramér's V for FLAG_PHONE: 0.021990638240084764

------------------------------------------------------------
Confusion matrix parfa FLAG_EMAIL con respeto a TARGET:
FLAG_EMAIL       0      1
TARGET                   
0           213165  12983
1            18773   1087
Cramér's V for FLAG_EMAIL: 0.0023645650771151965

------------------------------------------------------------
Confusion matrix parfa OCCUPATION_TYPE con respeto a TARGET:
OCCUPATION_TYPE  ACCOUNTANTS  CLEANING STAFF  COOKING STAFF  CORE STAFF  \
TARGET                                                                    
0                       7514            3362           4276       20671   
1                        379             372            507        1410   

OCCUPATION_TYPE  DRIVERS  HIGH SKILL TECH STAFF  HR STAFF  IT STAFF  LABORERS  \
TARGET                                                                          
0                  13195                   8584       436       389     39540   
1                   1699                    560        28        32      4652   

OCCUPATION_TYPE  LOW-SKILL LABORERS  MANAGERS  MEDICINE STAFF  \
TARGET                                                          
0                              1383     16048            6342   
1                               296      1092             454   

OCCUPATION_TYPE  PRIVATE SERVICE STAFF  REALTY AGENTS  SALES STAFF  \
TARGET                                                               
0                                 1992            549        23201   
1                                  136             44         2446   

OCCUPATION_TYPE  SECRETARIES  SECURITY STAFF  WAITERS/BARMEN STAFF  
TARGET                                                              
0                        993            4751                   959  
1                         74             585                   115  
Cramér's V for OCCUPATION_TYPE: 0.08084491419477519

------------------------------------------------------------
Confusion matrix parfa CNT_FAM_MEMBERS con respeto a TARGET:
CNT_FAM_MEMBERS   1.0     2.0    3.0    4.0   5.0   6.0   7.0   8.0   9.0   \
TARGET                                                                       
0                49807  117020  38380  18067  2503   288    57    10     5   
1                 4573    9611   3652   1702   262    47     5     5     0   

CNT_FAM_MEMBERS  10.0  11.0  12.0  13.0  14.0  15.0  16.0  20.0  
TARGET                                                           
0                   2     0     1     0     2     1     1     2  
1                   1     1     0     1     0     0     0     0  
Cramér's V for CNT_FAM_MEMBERS: 0.022443597931890227

------------------------------------------------------------
Confusion matrix parfa REGION_RATING_CLIENT con respeto a TARGET:
REGION_RATING_CLIENT      1       2      3
TARGET                                    
0                     24430  167304  34414
1                      1262   14307   4291
Cramér's V for REGION_RATING_CLIENT: 0.05797431452905184

------------------------------------------------------------
Confusion matrix parfa REGION_RATING_CLIENT_W_CITY con respeto a TARGET:
REGION_RATING_CLIENT_W_CITY      1       2      3
TARGET                                           
0                            25941  169054  31153
1                             1349   14489   4022
Cramér's V for REGION_RATING_CLIENT_W_CITY: 0.060540064376819976

------------------------------------------------------------
Confusion matrix parfa WEEKDAY_APPR_PROCESS_START con respeto a TARGET:
WEEKDAY_APPR_PROCESS_START  FRIDAY  MONDAY  SATURDAY  SUNDAY  THURSDAY  \
TARGET                                                                   
0                            36887   37472     24962   11869     37218   
1                             3281    3152      2141    1022      3305   

WEEKDAY_APPR_PROCESS_START  TUESDAY  WEDNESDAY  
TARGET                                          
0                             39534      38206  
1                              3580       3379  
Cramér's V for WEEKDAY_APPR_PROCESS_START: 0.004495454281500501

------------------------------------------------------------
Confusion matrix parfa REG_REGION_NOT_LIVE_REGION con respeto a TARGET:
REG_REGION_NOT_LIVE_REGION       0     1
TARGET                                  
0                           222748  3400
1                            19502   358
Cramér's V for REG_REGION_NOT_LIVE_REGION: 0.0062680653140641555

------------------------------------------------------------
Confusion matrix parfa REG_REGION_NOT_WORK_REGION con respeto a TARGET:
REG_REGION_NOT_WORK_REGION       0      1
TARGET                                   
0                           214743  11405
1                            18745   1115
Cramér's V for REG_REGION_NOT_WORK_REGION: 0.006750820541451695

------------------------------------------------------------
Confusion matrix parfa LIVE_REGION_NOT_WORK_REGION con respeto a TARGET:
LIVE_REGION_NOT_WORK_REGION       0     1
TARGET                                   
0                            216953  9195
1                             19015   845
Cramér's V for LIVE_REGION_NOT_WORK_REGION: 0.0015818927525979585

------------------------------------------------------------
Confusion matrix parfa REG_CITY_NOT_LIVE_CITY con respeto a TARGET:
REG_CITY_NOT_LIVE_CITY       0      1
TARGET                               
0                       209170  16978
1                        17481   2379
Cramér's V for REG_CITY_NOT_LIVE_CITY: 0.04516813958686674

------------------------------------------------------------
Confusion matrix parfa REG_CITY_NOT_WORK_CITY con respeto a TARGET:
REG_CITY_NOT_WORK_CITY       0      1
TARGET                               
0                       175412  50736
1                        13878   5982
Cramér's V for REG_CITY_NOT_WORK_CITY: 0.04965344090426648

------------------------------------------------------------
Confusion matrix parfa LIVE_CITY_NOT_WORK_CITY con respeto a TARGET:
LIVE_CITY_NOT_WORK_CITY       0      1
TARGET                                
0                        186403  39745
1                         15493   4367
Cramér's V for LIVE_CITY_NOT_WORK_CITY: 0.03126228732573633

------------------------------------------------------------
Confusion matrix parfa ORGANIZATION_TYPE con respeto a TARGET:
ORGANIZATION_TYPE  ADVERTISING  AGRICULTURE  BANK  BUSINESS ENTITY TYPE 1  \
TARGET                                                                      
0                          316         1761  1858                    4360   
1                           29          207   102                     394   

ORGANIZATION_TYPE  BUSINESS ENTITY TYPE 2  BUSINESS ENTITY TYPE 3  CLEANING  \
TARGET                                                                        
0                                    7733                   49490       193   
1                                     723                    5107        22   

ORGANIZATION_TYPE  CONSTRUCTION  CULTURE  ELECTRICITY  EMERGENCY  GOVERNMENT  \
TARGET                                                                         
0                          4787      291          716        418        7661   
1                           632       15           50         31         567   

ORGANIZATION_TYPE  HOTEL  HOUSING  INDUSTRY: TYPE 1  INDUSTRY: TYPE 10  \
TARGET                                                                   
0                    738     2196               731                 84   
1                     48      185                92                  3   

ORGANIZATION_TYPE  INDUSTRY: TYPE 11  INDUSTRY: TYPE 12  INDUSTRY: TYPE 13  \
TARGET                                                                       
0                               1959                291                 45   
1                                195                 12                  7   

ORGANIZATION_TYPE  INDUSTRY: TYPE 2  INDUSTRY: TYPE 3  INDUSTRY: TYPE 4  \
TARGET                                                                    
0                               338              2341               631   
1                                21               262                74   

ORGANIZATION_TYPE  INDUSTRY: TYPE 5  INDUSTRY: TYPE 6  INDUSTRY: TYPE 7  \
TARGET                                                                    
0                               457                81               972   
1                                32                 6                80   

ORGANIZATION_TYPE  INDUSTRY: TYPE 8  INDUSTRY: TYPE 9  INSURANCE  \
TARGET                                                             
0                                16              2535        454   
1                                 2               185         25   

ORGANIZATION_TYPE  KINDERGARTEN  LEGAL SERVICES  MEDICINE  MILITARY  MOBILE  \
TARGET                                                                        
0                          5157             229      8385      1974     228   
1                           383              22       590       110      22   

ORGANIZATION_TYPE  OTHER  POLICE  POSTAL  REALTOR  RELIGION  RESTAURANT  \
TARGET                                                                    
0                  12335    1794    1568      285        61        1295   
1                   1030      99     141       36         3         166   

ORGANIZATION_TYPE  SCHOOL  SECURITY  SECURITY MINISTRIES  SELF-EMPLOYED  \
TARGET                                                                    
0                    6684      2305                 1494          27621   
1                     421       273                   68           3105   

ORGANIZATION_TYPE  SERVICES  TELECOM  TRADE: TYPE 1  TRADE: TYPE 2  \
TARGET                                                               
0                      1163      421            243           1419   
1                        80       37             27            104   

ORGANIZATION_TYPE  TRADE: TYPE 3  TRADE: TYPE 4  TRADE: TYPE 5  TRADE: TYPE 6  \
TARGET                                                                          
0                           2561             52             38            464   
1                            295              0              3             19   

ORGANIZATION_TYPE  TRADE: TYPE 7  TRANSPORT: TYPE 1  TRANSPORT: TYPE 2  \
TARGET                                                                   
0                           5670                157               1619   
1                            583                  7                140   

ORGANIZATION_TYPE  TRANSPORT: TYPE 3  TRANSPORT: TYPE 4  UNIVERSITY    XNA  
TARGET                                                                      
0                                800               3903        1001  41769  
1                                154                403          59   2372  
Cramér's V for ORGANIZATION_TYPE: 0.07154282732719068

------------------------------------------------------------
Confusion matrix parfa FONDKAPREMONT_MODE con respeto a TARGET:
FONDKAPREMONT_MODE  NOT SPECIFIED  ORG SPEC ACCOUNT  REG OPER ACCOUNT  \
TARGET                                                                  
0                            4183              4231             54982   
1                             348               264              4146   

FONDKAPREMONT_MODE  REG OPER SPEC ACCOUNT  
TARGET                                     
0                                    9040  
1                                     623  
Cramér's V for FONDKAPREMONT_MODE: 0.012850572159249297

------------------------------------------------------------
Confusion matrix parfa HOUSETYPE_MODE con respeto a TARGET:
HOUSETYPE_MODE  BLOCK OF FLATS  SPECIFIC HOUSING  TERRACED HOUSE
TARGET                                                          
0                       112116              1056             869
1                         8389               125              86
Cramér's V for HOUSETYPE_MODE: 0.014947668199276329

------------------------------------------------------------
Confusion matrix parfa WALLSMATERIAL_MODE con respeto a TARGET:
WALLSMATERIAL_MODE  BLOCK  MIXED  MONOLITHIC  OTHERS  PANEL  STONE, BRICK  \
TARGET                                                                      
0                    6902   1659        1355    1188  49517         48030   
1                     526    140          65     111   3368          3849   

WALLSMATERIAL_MODE  WOODEN  
TARGET                      
0                     3875  
1                      418  
Cramér's V for WALLSMATERIAL_MODE: 0.03001729551410803

------------------------------------------------------------
Confusion matrix parfa EMERGENCYSTATE_MODE con respeto a TARGET:
EMERGENCYSTATE_MODE      NO   YES
TARGET                           
0                    118703  1699
1                      8913   180
Cramér's V for EMERGENCYSTATE_MODE: 0.01169416294243119

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_2 con respeto a TARGET:
FLAG_DOCUMENT_2       0  1
TARGET                    
0                226139  9
1                 19857  3
Cramér's V for FLAG_DOCUMENT_2: 0.0025764726142198224

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_3 con respeto a TARGET:
FLAG_DOCUMENT_3      0       1
TARGET                        
0                66692  159456
1                 4371   15489
Cramér's V for FLAG_DOCUMENT_3: 0.0449053947444949

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_4 con respeto a TARGET:
FLAG_DOCUMENT_4       0   1
TARGET                     
0                226128  20
1                 19860   0
Cramér's V for FLAG_DOCUMENT_4: 0.0

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_5 con respeto a TARGET:
FLAG_DOCUMENT_5       0     1
TARGET                       
0                222737  3411
1                 19568   292
Cramér's V for FLAG_DOCUMENT_5: 0.0

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_6 con respeto a TARGET:
FLAG_DOCUMENT_6       0      1
TARGET                        
0                205844  20304
1                 18673   1187
Cramér's V for FLAG_DOCUMENT_6: 0.028860072636401262

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_7 con respeto a TARGET:
FLAG_DOCUMENT_7       0   1
TARGET                     
0                226106  42
1                 19858   2
Cramér's V for FLAG_DOCUMENT_7: 0.0

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_8 con respeto a TARGET:
FLAG_DOCUMENT_8       0      1
TARGET                        
0                207586  18562
1                 18396   1464
Cramér's V for FLAG_DOCUMENT_8: 0.008055564379763382

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_9 con respeto a TARGET:
FLAG_DOCUMENT_9       0    1
TARGET                      
0                225254  894
1                 19801   59
Cramér's V for FLAG_DOCUMENT_9: 0.003670711674578854

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_10 con respeto a TARGET:
FLAG_DOCUMENT_10       0  1
TARGET                     
0                 226141  7
1                  19860  0
Cramér's V for FLAG_DOCUMENT_10: 0.0

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_11 con respeto a TARGET:
FLAG_DOCUMENT_11       0    1
TARGET                       
0                 225250  898
1                  19800   60
Cramér's V for FLAG_DOCUMENT_11: 0.003494245356856063

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_12 con respeto a TARGET:
FLAG_DOCUMENT_12       0  1
TARGET                     
0                 226146  2
1                  19860  0
Cramér's V for FLAG_DOCUMENT_12: 0.0

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_13 con respeto a TARGET:
FLAG_DOCUMENT_13       0    1
TARGET                       
0                 225280  868
1                  19832   28
Cramér's V for FLAG_DOCUMENT_13: 0.010668706807718871

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_14 con respeto a TARGET:
FLAG_DOCUMENT_14       0    1
TARGET                       
0                 225463  685
1                  19836   24
Cramér's V for FLAG_DOCUMENT_14: 0.008886516785392559

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_15 con respeto a TARGET:
FLAG_DOCUMENT_15       0    1
TARGET                       
0                 225858  290
1                  19851    9
Cramér's V for FLAG_DOCUMENT_15: 0.0059359632723035675

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_16 con respeto a TARGET:
FLAG_DOCUMENT_16       0     1
TARGET                        
0                 223848  2300
1                  19743   117
Cramér's V for FLAG_DOCUMENT_16: 0.01156868111186073

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_17 con respeto a TARGET:
FLAG_DOCUMENT_17       0   1
TARGET                      
0                 226086  62
1                  19858   2
Cramér's V for FLAG_DOCUMENT_17: 0.0014222099477547469

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_18 con respeto a TARGET:
FLAG_DOCUMENT_18       0     1
TARGET                        
0                 224229  1919
1                  19746   114
Cramér's V for FLAG_DOCUMENT_18: 0.007926583281735216

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_19 con respeto a TARGET:
FLAG_DOCUMENT_19       0    1
TARGET                       
0                 226002  146
1                  19850   10
Cramér's V for FLAG_DOCUMENT_19: 0.0

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_20 con respeto a TARGET:
FLAG_DOCUMENT_20       0    1
TARGET                       
0                 226032  116
1                  19848   12
Cramér's V for FLAG_DOCUMENT_20: 0.0

------------------------------------------------------------
Confusion matrix parfa FLAG_DOCUMENT_21 con respeto a TARGET:
FLAG_DOCUMENT_21       0   1
TARGET                      
0                 226079  69
1                  19849  11
Cramér's V for FLAG_DOCUMENT_21: 0.0026689070623304367

------------------------------------------------------------
Confusion matrix parfa AMT_REQ_CREDIT_BUREAU_HOUR con respeto a TARGET:
AMT_REQ_CREDIT_BUREAU_HOUR     0.0   1.0  2.0  3.0  4.0
TARGET                                                 
0                           195041  1116   42    6    1
1                            16304   101    6    0    0
Cramér's V for AMT_REQ_CREDIT_BUREAU_HOUR: 0.0

------------------------------------------------------------
Confusion matrix parfa AMT_REQ_CREDIT_BUREAU_DAY con respeto a TARGET:
AMT_REQ_CREDIT_BUREAU_DAY     0.0  1.0  2.0  3.0  4.0  5.0  6.0  8.0  9.0
TARGET                                                                   
0                          195139  919   76   35   19    8    7    1    2
1                           16300   97   11    1    2    0    0    0    0
Cramér's V for AMT_REQ_CREDIT_BUREAU_DAY: 0.0034934942279904696

------------------------------------------------------------
Confusion matrix parfa AMT_REQ_CREDIT_BUREAU_WEEK con respeto a TARGET:
AMT_REQ_CREDIT_BUREAU_WEEK     0.0   1.0  2.0  3.0  4.0  5.0  6.0  7.0  8.0
TARGET                                                                     
0                           189909  6052  143   43   28    8   16    2    5
1                            15877   510   16    3    3    1    1    0    0
Cramér's V for AMT_REQ_CREDIT_BUREAU_WEEK: 0.0

------------------------------------------------------------
Confusion matrix parfa AMT_REQ_CREDIT_BUREAU_QRT con respeto a TARGET:
AMT_REQ_CREDIT_BUREAU_QRT    0.0    1.0    2.0   3.0   4.0   5.0   6.0   7.0   \
TARGET                                                                          
0                          158758  25235  10519  1279   345    41    18     5   
1                           13538   1768    952    95    46     3     7     1   

AMT_REQ_CREDIT_BUREAU_QRT  8.0   19.0  
TARGET                                 
0                             6     0  
1                             0     1  
Cramér's V for AMT_REQ_CREDIT_BUREAU_QRT: 0.020681870368418964

------------------------------------------------------------
Confusion matrix parfa TARGET con respeto a TARGET:
TARGET       0      1
TARGET               
0       226148      0
1            0  19860
Cramér's V for TARGET: 0.9999726127135284

En resumen, aunque algunas variables muestran des diferencias en las matrices de confusión, su capacidad para predecir el incumplimiento de pago parece limitada debido a los bajos valores del coeficiente de Cramer.

Las variables más discriminantes entre las clases de TARGET son FLAG_MOBIL, FLAG_EMP_PHONE, FLAG_WORK_PHONE, y variables regionales como REG_REGION_NOT_LIVE_REGION y REG_CITY_NOT_LIVE_CITY. Estas variables presentan grandes diferencias en sus distribuciones y podrían ser indicadores importantes para la predicción de TARGET.

El valor de Cramér's V mide la fuerza de la asociación entre dos variables categóricas, y varía entre 0 y 1. Algunas de las valores importantes obtenidos son

• ORGANIZATION_TYPE: 0.0715, una asociación bastante moderada (si comparamos con las otras variables) con la variable TARGET.
• Otras variables (como FLAG_DOCUMENT_2, FLAG_DOCUMENT_10, etc.) tienen valores de Cramér's V cercanos a cero, lo que indica una asociación muy débil o nula.

VALORES NULOS DE CATEGORICAS A TRATAR /

• Decidi coger las variables que a primer vista se pueden rellenar con la moda (el valor mas frecuente); porque a lo mejor son variables que son indispensables para tener un modelo preciso.
• Los documentos si son faltantes significa que probablemente no han sido entregados asi que decidi rellenar con 0s.
• Y agrupar las que tienen sentido estar rellenadas con 'sin valor' o 'desconocido'

#mis de columnas categóricas
categorical_vars = [
    'NAME_CONTRACT_TYPE', 'CODE_GENDER', 'FLAG_OWN_CAR', 'FLAG_OWN_REALTY',
    'NAME_TYPE_SUITE', 'NAME_INCOME_TYPE', 'NAME_EDUCATION_TYPE', 
    'NAME_FAMILY_STATUS', 'NAME_HOUSING_TYPE', 'OCCUPATION_TYPE', 
    'FONDKAPREMONT_MODE', 'HOUSETYPE_MODE', 'WALLSMATERIAL_MODE', 
    'EMERGENCYSTATE_MODE', 'ORGANIZATION_TYPE', 'FLAG_DOCUMENT_2', 
    'FLAG_DOCUMENT_3', 'FLAG_DOCUMENT_4', 'FLAG_DOCUMENT_5', 'FLAG_DOCUMENT_6', 
    'FLAG_DOCUMENT_7', 'FLAG_DOCUMENT_8', 'FLAG_DOCUMENT_9', 'FLAG_DOCUMENT_10', 
    'FLAG_DOCUMENT_11', 'FLAG_DOCUMENT_12', 'FLAG_DOCUMENT_13', 'FLAG_DOCUMENT_14', 
    'FLAG_DOCUMENT_15', 'FLAG_DOCUMENT_16', 'FLAG_DOCUMENT_17', 'FLAG_DOCUMENT_18', 
    'FLAG_DOCUMENT_19', 'FLAG_DOCUMENT_20', 'FLAG_DOCUMENT_21', 
    'AMT_REQ_CREDIT_BUREAU_HOUR', 'AMT_REQ_CREDIT_BUREAU_DAY', 
    'AMT_REQ_CREDIT_BUREAU_WEEK', 'AMT_REQ_CREDIT_BUREAU_QRT'
]

#les divido segun la imputacion que decidi seguir:
fill_with_most_frequent = [
    'NAME_CONTRACT_TYPE', 'CODE_GENDER', 'FLAG_OWN_CAR', 'FLAG_OWN_REALTY', 
    'NAME_INCOME_TYPE', 'NAME_EDUCATION_TYPE', 'NAME_FAMILY_STATUS', 
    'ORGANIZATION_TYPE', 'FONDKAPREMONT_MODE', 'HOUSETYPE_MODE', 'OCCUPATION_TYPE'
]

fill_with_unknown = [
    'NAME_TYPE_SUITE', 'NAME_HOUSING_TYPE', 'FONDKAPREMONT_MODE', 
    'HOUSETYPE_MODE', 'WALLSMATERIAL_MODE', 'EMERGENCYSTATE_MODE', 
    'AMT_REQ_CREDIT_BUREAU_HOUR', 'AMT_REQ_CREDIT_BUREAU_DAY', 
    'AMT_REQ_CREDIT_BUREAU_WEEK', 'AMT_REQ_CREDIT_BUREAU_QRT'
]

fill_with_zero = [
    'FLAG_DOCUMENT_2', 'FLAG_DOCUMENT_3', 'FLAG_DOCUMENT_4', 'FLAG_DOCUMENT_5', 
    'FLAG_DOCUMENT_6', 'FLAG_DOCUMENT_7', 'FLAG_DOCUMENT_8', 'FLAG_DOCUMENT_9', 
    'FLAG_DOCUMENT_10', 'FLAG_DOCUMENT_11', 'FLAG_DOCUMENT_12', 'FLAG_DOCUMENT_13', 
    'FLAG_DOCUMENT_14', 'FLAG_DOCUMENT_15', 'FLAG_DOCUMENT_16', 'FLAG_DOCUMENT_17', 
    'FLAG_DOCUMENT_18', 'FLAG_DOCUMENT_19', 'FLAG_DOCUMENT_20', 'FLAG_DOCUMENT_21'
]

#imputar 
#moda
imputer = SimpleImputer(strategy='most_frequent')

data_train_input[fill_with_most_frequent] = imputer.fit_transform(data_train_input[fill_with_most_frequent])
data_test_input[fill_with_most_frequent] = imputer.transform(data_test_input[fill_with_most_frequent])

#unknown
for col in fill_with_unknown:
    if col in data_train_input.columns:
        # Verificar si "Desconocido" ya está como categoría antes de añadirla
        if data_train_input[col].dtype.name == "category" and "Desconocido" not in data_train_input[col].cat.categories:
            data_train_input[col] = data_train_input[col].cat.add_categories(["Desconocido"])
        if data_test_input[col].dtype.name == "category" and "Desconocido" not in data_test_input[col].cat.categories:
            data_test_input[col] = data_test_input[col].cat.add_categories(["Desconocido"])

data_train_input[fill_with_unknown] = data_train_input[fill_with_unknown].fillna("Desconocido")
data_test_input[fill_with_unknown] = data_test_input[fill_with_unknown].fillna("Desconocido")

#0s
data_train_input[fill_with_zero] = data_train_input[fill_with_zero].fillna(0)
data_test_input[fill_with_zero] = data_test_input[fill_with_zero].fillna(0)

# Verificación
print("Valores faltantes después de la imputación en data_train_input:")
print(data_train_input[categorical_vars].isnull().sum())

print("\nValores faltantes después de la imputación en data_test_input:")
print(data_test_input[categorical_vars].isnull().sum())

Valores faltantes después de la imputación en data_train_input:
NAME_CONTRACT_TYPE            0
CODE_GENDER                   0
FLAG_OWN_CAR                  0
FLAG_OWN_REALTY               0
NAME_TYPE_SUITE               0
NAME_INCOME_TYPE              0
NAME_EDUCATION_TYPE           0
NAME_FAMILY_STATUS            0
NAME_HOUSING_TYPE             0
OCCUPATION_TYPE               0
FONDKAPREMONT_MODE            0
HOUSETYPE_MODE                0
WALLSMATERIAL_MODE            0
EMERGENCYSTATE_MODE           0
ORGANIZATION_TYPE             0
FLAG_DOCUMENT_2               0
FLAG_DOCUMENT_3               0
FLAG_DOCUMENT_4               0
FLAG_DOCUMENT_5               0
FLAG_DOCUMENT_6               0
FLAG_DOCUMENT_7               0
FLAG_DOCUMENT_8               0
FLAG_DOCUMENT_9               0
FLAG_DOCUMENT_10              0
FLAG_DOCUMENT_11              0
FLAG_DOCUMENT_12              0
FLAG_DOCUMENT_13              0
FLAG_DOCUMENT_14              0
FLAG_DOCUMENT_15              0
FLAG_DOCUMENT_16              0
FLAG_DOCUMENT_17              0
FLAG_DOCUMENT_18              0
FLAG_DOCUMENT_19              0
FLAG_DOCUMENT_20              0
FLAG_DOCUMENT_21              0
AMT_REQ_CREDIT_BUREAU_HOUR    0
AMT_REQ_CREDIT_BUREAU_DAY     0
AMT_REQ_CREDIT_BUREAU_WEEK    0
AMT_REQ_CREDIT_BUREAU_QRT     0
dtype: int64

Valores faltantes después de la imputación en data_test_input:
NAME_CONTRACT_TYPE            0
CODE_GENDER                   0
FLAG_OWN_CAR                  0
FLAG_OWN_REALTY               0
NAME_TYPE_SUITE               0
NAME_INCOME_TYPE              0
NAME_EDUCATION_TYPE           0
NAME_FAMILY_STATUS            0
NAME_HOUSING_TYPE             0
OCCUPATION_TYPE               0
FONDKAPREMONT_MODE            0
HOUSETYPE_MODE                0
WALLSMATERIAL_MODE            0
EMERGENCYSTATE_MODE           0
ORGANIZATION_TYPE             0
FLAG_DOCUMENT_2               0
FLAG_DOCUMENT_3               0
FLAG_DOCUMENT_4               0
FLAG_DOCUMENT_5               0
FLAG_DOCUMENT_6               0
FLAG_DOCUMENT_7               0
FLAG_DOCUMENT_8               0
FLAG_DOCUMENT_9               0
FLAG_DOCUMENT_10              0
FLAG_DOCUMENT_11              0
FLAG_DOCUMENT_12              0
FLAG_DOCUMENT_13              0
FLAG_DOCUMENT_14              0
FLAG_DOCUMENT_15              0
FLAG_DOCUMENT_16              0
FLAG_DOCUMENT_17              0
FLAG_DOCUMENT_18              0
FLAG_DOCUMENT_19              0
FLAG_DOCUMENT_20              0
FLAG_DOCUMENT_21              0
AMT_REQ_CREDIT_BUREAU_HOUR    0
AMT_REQ_CREDIT_BUREAU_DAY     0
AMT_REQ_CREDIT_BUREAU_WEEK    0
AMT_REQ_CREDIT_BUREAU_QRT     0
dtype: int64

Guardado de la tabla

# Rutas para guardar los archivos
train_output_path = "../data/processed_data/processed_data_input/data_train_preprocessing_missing_outlier.csv"
test_output_path = "../data/processed_data/processed_data_input/data_test_preprocessing_missing_outlier.csv"

# Crear las carpetas (si no existen todavia)
os.makedirs(os.path.dirname(train_output_path), exist_ok=True)
os.makedirs(os.path.dirname(test_output_path), exist_ok=True)

# Guardar los DataFrames como .CSV
data_train_input.to_csv(train_output_path, index=False)
data_test_input.to_csv(test_output_path, index=False)

print(f"Archivo de entrenamiento guardado en: {train_output_path}")
print(f"Archivo de prueba guardado en: {test_output_path}")

Archivo de entrenamiento guardado en: ../data/processed_data/processed_data_input/data_train_preprocessing_missing_outlier.csv
Archivo de prueba guardado en: ../data/processed_data/processed_data_input/data_test_preprocessing_missing_outlier.csv

print(data_train_input.shape, data_test_input.shape)

(246008, 118) (61503, 118)